View
1
Download
0
Category
Preview:
Citation preview
저 시-비 리- 경 지 2.0 한민
는 아래 조건 르는 경 에 한하여 게
l 저 물 복제, 포, 전송, 전시, 공연 송할 수 습니다.
다 과 같 조건 라야 합니다:
l 하는, 저 물 나 포 경 , 저 물에 적 된 허락조건 명확하게 나타내어야 합니다.
l 저 터 허가를 면 러한 조건들 적 되지 않습니다.
저 에 른 리는 내 에 하여 향 지 않습니다.
것 허락규약(Legal Code) 해하 쉽게 약한 것 니다.
Disclaimer
저 시. 하는 원저 를 시하여야 합니다.
비 리. 하는 저 물 리 목적 할 수 없습니다.
경 지. 하는 저 물 개 , 형 또는 가공할 수 없습니다.
의학박사 학위논문
Clinically Relevant Biomedical Factors for Design & Development of
Practical Upper Limb Exoskeleton Rehabilitation Robots
실용적인 상지 외골격 재활 로봇
설계 및 개발을 위한 임상 적합성 기반 의공학적 인자
2018년 8월
서울대학교 대학원
의학과 의공학 전공
남 형 석
i
ii
i
ABSTRACT
Clinically Relevant Biomedical Factors for Design & Development of Practical Upper Limb
Exoskeleton Rehabilitation Robots
Hyung Seok Nam
Department of Biomedical Engineering,
Seoul National University College of Medicine
Introduction: There has been rapid growth in both the development and clinical
application of rehabilitation robots in the past decade. However, the goal of providing
maximal task-specific repetition of the limb movements to facilitate neuroplasticity
and functional recovery in neurorehabilitation, which is significantly superior to
conventional rehabilitation therapies, has not yet been achieved. The aim of this study
is to identify clinically relevant biomedical factors, distinguishable from simple
biomechanical factors, for the design and development of practical but simple
neurorehabilitation robots, focusing on exoskeleton-type robots.
Methods: A demand survey was performed on 48 potential users with stroke or
neuromuscular diseases to identify the patients’ practical needs, which may serve as
a goal for rehabilitation therapy. As spasticity is a common problem when applying
rehabilitation robots to patients with central nervous system disorders, biomechanical
response to spasticity was evaluated in 20 chronic stroke patients with various grades
of spasticity to characterize the spasticity induced resistance and to determine the
minimal torque output required for motors in major robot joints. An inertial
ii
measurement unit (IMU) sensor based motion capture system was used to determine
workspace and range of motion (ROM) for major upper extremity joints in ten healthy
subjects, while performing the Action Research Arm Test (ARAT) and top ranked
activities of daily living (ADLs) from the demand survey. The same evaluation
method was applied to nine stroke patients with Brunnstrom stages ranging from 3 to
6 to identify the characteristics of patient movements and stroke recovery patterns.
For user-intent driven control, an image-processing based robot control system was
proposed and a prototype for a hand rehabilitation robot was developed. A usability
study was performed with physicians, engineers, therapists, and stroke patients to
evaluate the robot’s clinical feasibility.
Results: In the demand survey, handling foods, dressing, and moving close items
were highly necessary ADL functions for both exoskeleton and external robot arm
types. Stroke patients demonstrated high demand for self-exercise with exoskeleton.
The maximal resistance torques caused by low (modified Ashworth scale (MAS) 0,
1), intermediate (MAS 1+), and high (MAS 2 and 3) grade spasticity were 3.68 ± 2.42,
5.94 ± 2.55, and 8.25 ± 3.35 Nm for the elbow flexor (p < 0.001, between each grade)
and 4.23 ± 1.75, 5.68 ± 1.96, and 5.44 ± 2.02 Nm for the wrist flexor (p < 0.001, for
low versus intermediate, low versus high grade spasticity). In healthy subjects, the
size of the workspace during the ARAT tasks was 0.53 m (x-axis, left-right) × 0.92 m
(y-axis, front-back) × 0.89 m (z-axis, up-down) for the dominant hand. For ADL tasks,
the workspace size was 0.71 m × 0.70 m × 0.86 m for the dominant hand which was
significantly larger than the non-dominant hand (p ≤ 0.011). The ROM for major
iii
joints of the upper extremity during the ARAT tasks were 109.15 ± 18.82° (elbow
flexion / extension), 105.23 ± 15.38° (forearm supination / pronation), 91.99 ± 20.98°
(shoulder internal / external rotation), and 82.90 ± 22.52° (wrist dorsiflexion /
volarflexion), whereas the corresponding ROM for the dominant side during the ADL
tasks were 120.61 ± 23.64°, 128.09 ± 22.04°, 111.56 ± 31.88°, and 113.70 ± 18.26°,
respectively. Of the parameters that showed significant differences in values between
healthy subjects and patients and also significant correlation with clinical measures,
the average amplitude of the forearm supination / pronation angle during the ARAT
domain 4 tasks demonstrated the greatest decline in severely impaired patients
compared to normal subjects (29.83%) and also largest difference between severely
and mildly impaired patients (48.46%). For the usability test for the image processing
based user-intent driven hand rehabilitation robot, the participants found the device
interesting (5.7 ± 1.2), motivating (5.8 ± 0.9), and as having less possibility of injury
or safety issues (6.1 ± 1.1); however, the levels of difficulty (4.8 ± 1.9) and comfort
(4.9 ± 1.3) were relatively low.
Conclusions: The results of this research will serve as a basis for the design and
development of a practical and portable but clinically relevant neurorehabilitation
exoskeleton robot.
Keywords: Neurorehabilitation Robot; Stroke; Upper Extremity; Biomedical Factor;
Exoskeleton
Student Number: 2012-21741
iv
TABLE OF CONTENTS
Abstract ......................................................................................................... i
Table of Contents ....................................................................................... iv
List of Tables ............................................................................................. vii
List of Figures ........................................................................................... viii
List of Abbreviations ................................................................................... x
1. Introduction ............................................................................................... 1
1.1 Research on Upper Limb Exoskeleton Robots ................................. 1
1.2 Previous Robot Development and its Lessons .................................. 3
1.2.1 Clinical Application Experience with Two-axis Mirror Robot ................................................................................................. 3
1.2.2 Development Experience with Multi-axis Upper Extremity Exoskeleton ............................................................................. 3
1.2.3 Non-invasive Brain Machine Interface Control Methods ...... 7
1.3 Potential Factors for Investigation .................................................... 8
1.3.1 Robot Function (Purpose of the robot) ................................... 8
1.3.2 Robot Structure (Resistance and range of motion) ................. 9
1.3.3 Robot Control Method .................................................................................... 12
1.4 Objectives ....................................................................................... 13
2. Methods .................................................................................................... 14
v
2.1 Demand Investigation for Upper Limb Exoskeleton and Brain- Machine Interface on potential users (patients) ............................. 14
2.2 Biomedical Factor Investigation ..................................................... 18
2.2.1 Biomechanical Response of Exoskeleton in Spastic Elbows and Wrists ......................................................................... 18
2.2.2 Upper Limb Motion Characterization in Major Movements & Tasks ................................................................................. 25
2.3 Feasibility Study for User-intent Driven Robot Control Methods .. 30
2.3.1 Image-processing Based Control and its Feasibility ............ 30
2.3.1.1 Development of Image-processing Based Hand Rehabilitation Robot .................................................... 30
2.3.1.2 Preliminary Usability Test ............................................................... 34
2.4 Statistical Analysis and Study Approval ......................................... 35
3. Results ...................................................................................................... 36
3.1 Practical Robot Functions in Demand ............................................ 36
3.2 Minimum Requirements for Motor Power in Major Joints to Overcome Spasticity ...................................................................... 48
3.3 Range of Motion and Movement Characteristics in Major Movements & Tasks ...................................................................... 57
3.3.1 Healthy Subjects ................................................................... 57
3.3.2 Stroke Patients ...................................................................... 63
3.4 Preliminary Usability Test for Image-processing Based Hand Rehabilitation Robot ...................................................................... 66
4. Discussion................................................................................................. 68
4.1 Demand survey for potential users of robots .................................. 68
vi
4.2 Biomechanical Response of Exoskeleton to Spasticity .................... 71
4.3 Kinematic Characteristics of Upper Extremity in Healthy Subjects and Stroke Patients ........................................................................ 76
4.4 Usability Test for an Image-processing Based Hand Rehabilitation Robot ............................................................................................. 82
4.5 Optimization of Neurorehabilitation Robot Design Regarding Clinical Settings ............................................................................. 84
4.6 Limitations ...................................................................................... 90
5. Conclusions and Future Work ............................................................... 93
Acknowledgments ..................................................................................... 95
Funding ...................................................................................................... 96
References .................................................................................................. 97
Supplemental Materials .......................................................................... 108
Appendix ................................................................................................... 115
국문초록 .................................................................................................... 116
vii
LIST OF TABLES
Table 1. Demographic Data of Demand Survey Participants ............................... 36
Table 2. Dependency, Importance, and Necessity for Assistive Robots in Unilateral
Impairment Group ........................................................................................... 40
Table 3. Dependency, Importance, and Necessity for Assistive Robots in Bilateral
Impairment Group ........................................................................................... 45
Table 4. Demographic Data of Enrolled Stroke Patients for Spasticity Resistance
Evaluation ....................................................................................................... 49
Table 5. Resistance Torque Values to Spastic Elbow and Wrist Joints.................. 53
Table 6. Coefficient of Variation during Major Movements for IMU-based Upper
Extremity Motion Capture System .................................................................... 58
Table 7. Range of Motion Angle between Right and Left Upper Extremities During
ARAT and ADL Tasks...................................................................................... 60
Table 8. Major Joint Angle Position and Change During Grasping / Pinching and
Reaching ......................................................................................................... 62
Table 9. Average Amplitude Angles and Acceleration for Significantly Declined
Parameters in Stroke Patients by Brunnstrom Stage ........................................... 64
viii
LIST OF FIGURES
Figure 1. A 6-axis semi-wearable upper limb exoskeleton prototype ..................... 5
Figure 2. A 4-axis upper limb wearable exoskeleton prototype ............................. 6
Figure 3. Schematic structure of the two-axis rehabilitation robot for resistant torque
measurement ................................................................................................... 19
Figure 4. A representative graph of the outcome parameters during resistant torque
measurement ................................................................................................... 23
Figure 5. IMU-based upper extremity motion capture system ............................. 26
Figure 6. Initial design for image-processing based user-intent driven 2-axis hand
rehabilitation robot .......................................................................................... 31
Figure 7. Main concept of the image-processing based user-intent driven hand
rehabilitation robot .......................................................................................... 33
Figure 8. Importance and necessity of external robotic arm for each activity of daily
living .............................................................................................................. 41
Figure 9. Importance and necessity of upper limb exoskeleton for each activity of
daily living ...................................................................................................... 42
Figure 10. Potential user needs for each activity of daily living by robot type and
laterality of impairment .................................................................................... 47
ix
Figure 11. Box-and-whisker plots for maximal resistance torque of elbow flexor,
elbow extensor, and wrist flexor ..................................................................... 54
Figure 12. Maximum resistance torque and catch angle by angular velocity for
elbow flexor spasticity ..................................................................................... 56
Figure 13. Average amplitude angles of joint movement segments during ARAT
tasks by Brunnstrom stage ................................................................................ 65
Figure 14. Usability test results for image-processing based hand rehabilitation
robot ............................................................................................................... 67
Figure 15. Improvement in Fugl-Meyer upper extremity motor score after robot-
assisted treatment ............................................................................................ 85
x
LIST OF ABBREVIATIONS
ADL Activity of daily living
ARAT Action Research Arm Test
BIG Bilateral impairment group
BMI Brain-machine interface
CNS Central nervous system
CoV Coefficient of Variance
CPP Catch point percentage
EEG Electroencephalogram
EMG Electromyography
F-M Fugl-Meyer
IMU Inertial measurement unit
MAS Modified Ashworth scale
MRI Magnetic resonance imaging
RMSE Root mean square error
ROM Range of motion
UIG Unilateral impairment group
1
1. INTRODUCTION
1.1 Research on Upper Limb Exoskeleton Robots
Because of the high incidence of stroke and the recent trend toward developing
rehabilitation robots, many types of rehabilitation robots for stroke rehabilitation have
been developed [1, 2]. Over the past few years, there have been considerable
improvements in neurorehabilitation robotics. However, not many types of
neurorehabilitation robots have entered the developmental stage for large-scale
randomized controlled clinical trials, nor have they been widely commercialized. Part
of the reason for this would be regulation issues pertaining to medical devices;
however, it seems the main reason is that those pre-developed rehabilitation robots
do not provide sufficient “task-specific high repetition” while appropriately
maintaining motivation [3]. From a clinical perspective, it is undeniable that more
task-specific repetition of the paralyzed extremity would lead to better recovery in
patients with limb paralysis caused by central nervous system (CNS) injuries or
disorders. In fact, most of the rehabilitation robots focus on providing high repetitions
unless they are developed for assistance.
In general, electromechanical devices and robots for upper limb rehabilitation are
classified into two categories: end-effector robots and exoskeletons [4]. Exoskeletons
have a structure in which the robot joints correspond to human joints [5]. These types
of robot are generally large and expensive, and they are usually fixed in space, which
makes them only usable in occupational therapy rooms in hospitals. Armeo® series
2
exoskeleton robots are the most widely used exoskeletons in clinics. End effector type
robots are relatively simple in structure. These robots usually let the patients hold a
handle with their hands and the handle generates power according to its trajectory and
direction. The joint of the robot does not correspond with human anatomical structure;
therefore, there are various types of end-effector robots [5]. InMotionTM is the most
representative type of the end-effector type robot. End-effector robots are generally
not in a wearable form; therefore, they are mostly fixed in one space and are mainly
used in the clinics. While it is not clear whether either type of rehabilitation robot is
better than another, numerous systematic reviews and meta-analyses showed
conflicting results on the efficacy of robot-assisted arm rehabilitation [2, 4]. There is
growing evidence that robot-assisted training improves both muscle strength and
functional abilities, however, whether the amount of increase in outcome measures is
clinically significant remains questionable, especially when considering cost-
effectiveness.
Regardless of the robot type, the most important reason that the efficacy of robot-
assisted training is currently not more prevalent or desired is that it simply does not
provide sufficient amount of task-specific repetition. Many factors contribute to this
limitation including patient’s medical status, functional status, socioeconomic factor,
hospital accessibility, insurance policies, etc. Because all these factors are not easily
controllable in many aspects, it is necessary to develop a rehabilitation robot that is
affordable, clinically feasible, and portable, which will make the robot more
accessible, and eventually maximize the task-specific repetition of the paralyzed limb.
3
1.2 Previous Robot Development and its Lessons
1.2.1 Clinical Application Experience with a Two-axis Mirror Robot
Several years ago, we developed a two-axis upper limb robot for robotic mirror
therapy in stroke patients [6]. This device was operated based on multiple inertial
measurement unit (IMU) sensors, which reflect the movement of the intact arm and
actuates the robotic arm on the hemiplegic side, while the patient is looking at the
mirror to provide an illusion that the paretic arm is really moving naturally. During
the development and preliminary clinical trial of the robotic mirror therapy device,
the robot could not actuate the elbow joint for patients with spasticity at grade 2 on
the modified Ashworth Scale (MAS) even though a pilot test was successfully
performed for healthy subjects and some selected patients. MAS grade 2 corresponds
to a marked increase of spasticity; however, many stroke patients have spastic joints
at greater than MAS grade 2. After the failure, the robot was equipped with motors
with greater torques and the clinical study showed beneficial effects regarding
proprioception, spatial hemineglect, and neuroplasticity supported by functional
magnetic resonance imaging (MRI) findings [7].
1.2.2 Development Experience with Multi-axis Upper Extremity Exoskeleton
In the beginning stage of our rehabilitation robot research, we designed and
developed a seven-axis upper extremity exoskeleton to expand the application of
robotic mirror therapy to the whole upper extremity (Fig. 1). Validation for the exact
4
reflection of the intact arm to the robotic arm through multiple IMU sensors was
performed. However, this development could not go further because of medical
device approval issues.
In addition to the medical device approval issues, the seven-axis robot was actually
not clinically feasible for use in clinics or at home due to its large size. Therefore, we
designed a second version of the exoskeleton with only four axes: shoulder internal /
external rotation, elbow flexion / extension, forearm supination / pronation, and wrist
dorsiflexion / volarflexion. Considering that most of the hemiplegic stroke patients
show earlier recovery in their proximal limb compared to their distal limb, we
eliminated the shoulder structure to reduce size and weight, and therefore provided
more portability. A desktop height-adjustable elbow support with movable wheels
was designed and manufactured to substitute and assist shoulder function. While this
device also requires the clearance of medical device related regulations, the most
appropriate user-intent driven control method remains unsolved and is under
development (Fig. 2).
5
Fig. 1. A 6-axis semi-wearable upper limb exoskeleton prototype is shown.
6
Fig. 2. A 4-axis upper limb wearable exoskeleton is shown. The weight of the robot is approximately 3.7kg.
7
1.2.3 Non-invasive Brain Machine Interface Control Methods
Most of the rehabilitation robots on the market are controlled via a monitor based
user interface, usually in the form of a game or virtual task. To control the robotic
arm by user-intent, commonly used methods are the electromyography (EMG) based
method and torque sensor based method. The concept itself is relatively simple, but
in reality, it is difficult to cancel out the noise and properly sense and actuate the robot,
and it is not usable in patients with complete paralysis.
Electroencephalography (EEG) based brain-machine interface (BMI) control
methods have been investigated over the years for robotic arm control. There are
some studies that report successful control of robots by EEG in a switch controller-
like manner; however, to the best of our knowledge, motor-imagery EEG based
control does not seem to be feasible for practical use. In our research team’s
preliminary studies, we attempted to record EEG signals while actually performing
the reaching task and also gain EEG signals from motor imagery [8]. After regression
learning, new EEG signals were given to control the external robotic arm, but the
decoded EEG did not show sufficient accuracy in terms of positional data. For the
completion of the robotic arm task, an image-guided compensation algorithm
regarding position and orientation was applied.
8
1.3 Potential Factors for Investigation
1.3.1 Robot Function (Purpose of the robot)
To focus on the accessibility and portability of the rehabilitation robot, it is important
to specify the main purpose of the robot. In general, rehabilitation robots are classified
into one of the two types depending on their purpose: robots for neurorehabilitation
purposes, which aim to enhance functional recovery of the limb by facilitating
neuroplasticity, and robots for assistive purposes, which focus on performing desired
tasks, usually activities of daily living (ADL). Robots for neurorehabilitation tend to
be either exoskeleton or end-effector type, and robots for assistance more likely have
the form of an external robotic arm, such as a feeding robot or the JACOTM robotic
arm. For neurorehabilitation, it is very essential to facilitate the repetitive feedback
loop consisting of brain – motor execution – motor actuation – visual and sensory
feedback to maximize neuroplasticity.
To the best of our knowledge, no specific robot has been commercially and widely
utilized to assist in performing ADLs that require the upper extremities, although
some are being used in limited circumstances [9]. Robots used only for treatment
purposes are fixed at a certain location in a hospital, and they do not require
portability in that setting. For the robots to be applicable in daily activities, they need
to be portable, simple, controllable according to the user’s intent, and be able to
involve real target objects instead of providing monitor screen or virtual reality. Most
of the robots use force sensors, torque sensors, or surface electromyography to
recognize user’s intent [10]. However, it is difficult for the people with severe limb
9
impairment to generate sufficient input signals for such sensors. Recently, active
research on BMI is being conducted to extract user intent directly from the brain
signals. However, it is very challenging to perform precise control of the robot with
electroencephalography signals, which is the most commonly used non-invasive
brain signal, because the signal-to-noise ratio is very low. Furthermore, BMI
technology involving invasive brain signals such as intracortical signals or
electrocorticography, is yet far from practical utilization [8, 11, 12]. Therefore, it is
important to clarify the purpose of the robot in the development and design stage, to
simplify and adapt the robot structure, including whether it needs to be in the form of
an external robotic arm or exoskeleton, so that it may be practically applied to ADL
with simple control modalities. In order to clarify the purpose, the practical demands
of the potential users with functional impairment need to be evaluated to avoid a
situation where a perfect robot for both control and function is developed but nobody
seeks to use it.
1.3.2 Robot Structure (Resistance and range of motion)
Robots for neurorehabilitation are mainly used by patients with CNS lesion or
disorders. The most common disease that requires neurorehabilitation robots is stroke.
In designing neurorehabilitation robots, understanding of the disease characteristics
is particularly important.
In minimizing the mass and complexity of the exoskeleton, spasticity may act as a
10
major challenge in neurorehabilitation robot design because the resistance induced
by spasticity may require higher torque output to actuate the joint than in a healthy
person. Therefore, the robot may require a motor with a higher output torque to be
able to perform the desired movements.
Spasticity is characterized by a velocity-dependent increase of resistance caused by
the exaggeration of the stretch reflex when the joint is passively stretched [13].
Spasticity occurs in upper motor neuron disorders including CNS lesions, and its
incidence ranges from 30 to 60% in stroke and 65 to 78% in spinal cord injury [14-
16]. Spasticity significantly affects sensorimotor function [17]. Especially in the
chronic stage, spastic joints commonly present non-reflex hypertonic features, which
mainly arise from soft tissue changes such as the shortening of the muscle and fibrotic
changes [13]. When non-reflex hypertonic features progress and predominate, they
may cause contracture or rigidity, which is represented as grade 4 in the MAS [18].
Non-reflex hypertonicity and rigidity are different from spasticity in that they are not
affected by passive velocity and direction [19]. In stroke, rigidity usually appears as
a sequelae of severe spasticity in the chronic stage. The target users for
neurorehabilitation robots among the stroke patients, especially for those developed
as assistive devices, would be patients in a chronic state, and a high percentage of
them demonstrate considerable spasticity. Therefore, spasticity would be an
important factor in designing neurorehabilitation robots to be widely and practically
used.
For a neurorehabilitation robot to perform a desired movement in the presence of up
11
to a certain degree of spasticity, the robot should have a motor with sufficient output
torque to overcome the resistance created by the spastic muscles [20], or it should
have a specific dynamic control algorithm to avoid high resistance during robot-
actuated movement; and to prevent overloading from mechanical or anatomical
compensation in adjacent joints [18]. Large fixed robots for treatment only in
rehabilitation facilities have high torque output motors; however, actuation with
excessive torque output may cause injury to spastic or rigid joints. Furthermore, in
order to develop a robot in a portable and wearable form for use in daily living, the
appropriate range of torque output requirements for the motors should be determined.
Some guidelines and measurements have been established for spastic joints; however,
to the best of our knowledge, definite evidence based on the measurements of patients
with spasticity is rare [20-22], and Alibiglou et al. [23] reported in 2008 from a study
on the spastic ankle joints of 20 stroke patients that quantitative measures and clinical
assessments lack significant correlation.
Since it is important to minimize the size and complexity of the neurorehabilitation
robot, the number of axes and the workspace of the robotic hand or the end-effector
should be minimized, but at the same time the robot needs to be able to perform
essential tasks in daily activities and functions. In the viewpoint of specific task
performance, human movement using the arm and actuation of the robot may seem
similar; however, when considering the mechanism of the performance, they are
actually significantly different. Moreover, it is possible to say that the biological and
engineering mechanisms are very opposite [24]. There have been numerous attempts
12
to analyze the 3D movements of the upper extremity during ADL using image based
motion analysis, IMU sensor based motion analysis, magnetic sensor systems, etc.
Positional and angular data during ADL such as drinking water, fastening a button,
touching the perineum, eating food, and combing hair were evaluated [25-27]. Chen
and Lum [28] performed a study with a spring operated exoskeleton to assess the
change in joint movements when assisted by the robot. It is important to have a
database on the position and joint angles while performing essential daily activities;
however, the movement patterns between healthy subjects and stroke patients differ
significantly, and the exoskeleton cannot be actuated exactly in the same manner as
the human limb. To optimize the design of the robot and to focus on the recovery of
specific functions, the dimensions of the essential workspace, range of motion (ROM)
of major joints of the upper extremity in the normal motion of healthy subjects, and
the characteristics of joint movements in stroke patients, all need to be evaluated.
1.3.3 Robot Control Method
To maximize neuroplasticity in using rehabilitation robots, the robot should move
according to the user’s intent, or at least the patient should be able to anticipate the
robot’s movement. Because EMG and torque sensor based controls are not applicable
to patients with flaccid paralysis or with only minimal volitional movements, in this
thesis we attempted to apply an image processing based approach that can reflect
user-intent. Visual compensation using camera images in the BMI control of the
robotic arm has recently been introduced [29]. Bang et al. [30] suggested an upper
13
limb rehabilitation robot system for precision control by image processing.
1.4 Objectives
The aim of the present study was to identify clinically relevant biomedical factors,
distinguishable with simple biomechanical factors, for the design and development
of practical but simple rehabilitation robots, focusing on exoskeleton type robots.
First, a survey was performed to assess potential users’ practical demands on robot
function. Biomechanical response to spasticity was evaluated with stroke patients
with various grades of spasticity to determine the minimal torque output of the robot
motor. The workspace and ROM for major joints in healthy subjects were evaluated
while performing the Action Research Arm Test (ARAT) and ADLs with high
demand. The characteristics of joint movement in stroke patients were identified.
Finally, based on determined biomedical factors, an image processing based user-
intent driven hand rehabilitation robot was developed, and a pilot usability test was
performed.
14
2. METHODS
2.1 Demand Investigation for Upper Limb Exoskeleton and Brain-
Machine Interface on potential users (patients)
A. Survey Development and ADL Items
For differentiation of utilizing external robotic arm and upper limb exoskeleton for
assistance, it was defined and explained to the survey participants that the external
robotic arm will assist as a helper external to the body performing the desired
movement or providing and placing the object that is needed for them to complete the
task. Exoskeleton was described that it would be in a completely wearable form to
assist the movements of the paralyzed part of the body, including the fingers in the
form of a glove. For control of the robot, the subjects were told to assume a perfect
BMI control system and a robot that could perform all of the tasks completely. The
subjects were informed about the three categories of BMI technology from non-
invasive to invasive to have a concept of BMI, however, for the survey they were
instructed to assume the best BMI system without significant risks. Regarding each
different task, it was assumed that the robot was specifically designed for the
designated ADL task, therefore, a robot for assisting eating and another robot for
assisting brushing teeth may have different forms of hardware and control
mechanisms regardless of total degree of freedom of the robot. It was also assumed
15
that a robot hand or a glove type exoskeleton with precise control is mounted on the
robot. They were instructed not to take the possibility of realization into consideration,
but rather focus on the importance and necessity of the ADL items themselves.
Fourteen ADLs were selected for rating in regards to both the external robotic arm
and upper limb exoskeleton, and 4 additional items were selected for exoskeletons.
The survey items were extracted from the commonly used ADL assessment tools,
such as the modified Barthel Index (MBI) [31] or Functional Independence Measure
(FIM) [32], as well as some results from our preliminary research. After the extraction,
a professional committee meeting consisting of 4 physiatrists, 2 rehabilitation robot
engineers, and 2 neuroscience engineers reviewed the survey items and selected
relevant survey items. The 14 selected ADL items were: 1) washing face, 2) brushing
teeth (including squeezing toothpaste), 3) hairdressing, 4) dressing (putting shirts on
and off), 5) eating, 6) handling foods (i.e. peeling a banana, opening a bottle cap, etc.),
7) cleaning (cleaning one’s desk), 8) moving close items, 9) smartphone (using a
smartphone or smart tablet), 10) computer (using a computer), 11) phone calls
(dialing and receiving a phone call), 12) writing, 13) switch control, and 14) purse
(putting in and taking out bills and cards from a purse / wallet). The 4 additional items
for exoskeletons were: 1) transfer (assisting bed to chair, chair to standing, etc.), 2)
toilet use, 3) self-exercise (of the upper extremity), and 4) wheelchair control (both
manual and electric). These items were not included for the external robotic arm
category they were considered not appropriate to be assisted by an external robot, and
the assistance method was clearly explained to the survey participants. The survey
16
form and detailed instructions are provided in the supplemental materials
(Supplement 1).
B. Participants
The survey was conducted on volunteers with severe functional impairments of either
unilateral or bilateral upper extremities caused by various neuromuscular diseases,
who are considered to be potential users of the external robotic arm or upper limb
exoskeleton controlled by a BMI system upon its development. A total of 48 patients
from the outpatient clinic of the Department of Rehabilitation in a tertiary hospital
volunteered to answer the survey. All participants provided written, informed consent
before enrollment.
C. Procedure
For each ADL selected, all participants were asked to rate their current level of
dependence on another person using a 5-point Likert scale (1: totally dependent, 2:
mostly dependent; 3: half dependent; 4: mostly independent, 5: totally independent).
Then, the participant rated the objective importance of the function from the
viewpoint of a developer based on their experience with severe functional impairment.
External robotic device and exoskeletons were rated separately, also using a 5-point
Likert scale (1: unimportant; 2: of little importance; 3: moderately important; 4:
important; 5: very important). The same items were presented to determine the
17
subjective necessity of the function from the viewpoint of a consumer based on their
current daily activities. They were asked if they would need to use it if a perfect robot
with the mentioned function was provided (1: not necessary, 2: of little necessity, 3:
moderately necessary, 4: necessary, 5: highly necessary). A short free interview was
done after the survey to assess detailed information on their answers.
D. Statistical Analysis
Descriptive statistical analyses were performed for the rating scores of each survey
items. Independent t tests for comparison of necessity between unilateral impaired
group and bilateral impaired group regarding each ADL items and type of robot were
performed. Paired t-tests were performed to compare the necessity of each item
between the two subgroups and also between the two types of robotic devices. A p
value less than 0.05 was considered statistically significant.
18
2.2 Biomedical Factor Investigation
2.2.1 Biomechanical Response of Exoskeleton in Spastic Elbows and Wrists
A. Robot Design and Settings
The robot used in this study was the 2-axis (elbow extension / flexion, wrist
dorsiflexion / volarflexion) planar upper limb exoskeleton robot system initially
designed and developed by our research team for a robotic mirror therapy system for
functional recovery of a hemiplegic arm [6]. This robot is equipped with two
brushless DC motors (EC45flat and EC90flat, Maxon Motor AG, Sachseln,
Switzerland) with sufficient torque outputs to overcome spasticity of the participants
and the robot is fixed to a metal frame desk. Torque sensors (TFF-500, CTAplus Co.,
Ltd., Daegu, South Korea; FT01-20NM, Forsentek Co., Ltd., Shenzhen, China) are
mounted on each joint to measure reaction torques between the human arm and the
robot actuators during the robotic therapy. For the real time monitoring platform,
LabVIEW® (National Instruments (NI) Corp., Austin, TX, USA) was used to control
and monitor the robotic movements in real time. The schematic structure of the
system and a photo of a participant during the trial are shown in Fig. 3.
19
Fig. 3. (A) Schematic structure of the system is shown. (B) A subject is equipped with a two-axis rehabilitation robot for
measurement of the resistance torque.
20
A. Participants
For evaluation, chronic stroke patients with a spastic upper limb were recruited.
Subjects matching all of the following inclusion criteria were included in the study:
1) spasticity of the hemiplegic upper limb with a MAS grade between 1 and 3 in the
elbow and / or the wrist flexor; 2) a history of stroke with more than 1 year since
onset; and 3) alert mental status sufficient for following instructions during the study.
The clinical experiment was performed between July and September in 2015. A total
of 20 chronic stroke patients and one healthy subject for comparison volunteered for
the study with written informed consent from the department of rehabilitation
medicine in two hospitals: Seoul National University Hospital (SNUH) and Seoul
National University Boramae Medical Center (SNUBMC).
B. Clinical Test Settings and Procedure
For each subject, two skilled rehabilitation specialists independently assessed the
spasticity of the elbow flexor, elbow extensor, and wrist flexor using the MAS scale.
Considering the variability of spasticity, the higher grade was used for classification.
After the subjects were equipped with the exoskeleton robot, the maximum flexion
and extension angle range for each joint was configured for each subject in order to
prevent injuries. The maximum flexion and extension angle were determined by a
clinical decision based on physical examinations performed by rehabilitation
specialists, and are the extent to which the joint angle cannot be further extended or
flexed due to severe spasticity, contracture, or pain. All subjects were given three
consecutive passive isokinetic movements back and forth by the robot within its ROM.
21
Three different angular velocities at the elbow joint (20, 40, and 60 ° /s) and six
different velocities at the wrist joint (20, 40, 60, 80, 100, and 120 ° /s) were applied.
The resistance torque at each joint and the corresponding angular position were
continuously recorded throughout the process. The test-retest reliability coefficients
of the torque sensor for the elbow and wrist were 0.977 and 0.931, respectively.
C. Extracted Parameters
In each trial for flexor spasticity, the maximal resistance torque during the extension,
the resistance torque value at the maximal flexion state (0%ROM), 1/3 point
(1/3ROM) and 2/3 point (2/3ROM) of the ROM, and full extension (100%ROM)
were calculated based on the measured torque data. For each trial of extensor
spasticity, the same parameters were extracted in the opposite direction, from full
extension to full flexion. The changing pattern of the stiffness for each trial was
computed and represented in graphs as dτ / dt, because all movements in a single trial
were isokinetic, making dƟ / dt a constant (Equation 1).
Stiffness (𝑁𝑁𝑁𝑁/𝑟𝑟𝑟𝑟𝑟𝑟) = 𝑟𝑟𝑑𝑑𝑟𝑟Ɵ
= 𝑟𝑟𝑑𝑑𝑟𝑟𝑑𝑑
× �𝑟𝑟Ɵ𝑟𝑟𝑑𝑑�−1
∝ 𝑘𝑘 𝑟𝑟𝑑𝑑𝑟𝑟𝑑𝑑
(𝑁𝑁𝑁𝑁/𝑠𝑠𝑠𝑠𝑠𝑠) (1)
In Equation 1, τ is the torque measured by the sensor, θ is the angular position, and
dτ / dt and dθ / dt are time derivatives of τ and θ. The k stands for a constant coefficient.
A sudden increase in the resistance torque during movement, where the time
22
derivative of the torque (dτ / dt) increase to a peak, was assumed as equivalent to a
“catch” phenomenon [33]. The angle at which the catch phenomenon occurs was
defined as the catch angle, and the catch angle (deg) was converted to catch point
percentage (CPP, %ROM) according to Equation 2, where the difference between
catch angle and initial angle is divided by total ROM. The CPP shows a low value
when the catch phenomenon occurs early during the passive ROM, and is 100% if
there is no catch.
Catch point percentage (%ROM)
= | catch angle (𝑟𝑟𝑠𝑠𝑑𝑑) − initial angle (𝑟𝑟𝑠𝑠𝑑𝑑)|
total ROM (𝑟𝑟𝑠𝑠𝑑𝑑) × 100 (2)
Throughout the study, the resistance torques caused by flexor spasticity during
actuation in the extension direction are given in positive values whereas the resistance
torques caused by extensor spasticity in the flexion direction are given in negative
values. A representative graph of the outcome parameters is shown in Fig. 4.
23
Fig. 4. A representative graph of the outcome parameters is shown. Blue lines indicate
resistance torque values and red lines show corresponding angular positions. The
catch phenomenon and the exponential decay of torque at the end of range of motion
are indicated.
24
D. Statistical Analysis
The degree of spasticity was categorized into three grades based on the MAS; low
(MAS grade 0, 1), intermediate (MAS grade 1+), and high grade (MAS grade 2, 3)
spasticity. Statistical analyses were performed for spasticity of the elbow flexor,
elbow extensor, and wrist flexor. The maximal resistance torque values for the three
spasticity grades over all trials were compared using the analysis of variance
(ANOVA) method followed by post-hoc analyses. Resistance torque values at 0%,
1/3, 2/3, and 100% ROM were compared using the repeated measures ANOVA (RM-
ANOVA) method with ROM position as within-subjects factor and spasticity grades
as between-subject factor, followed by post hoc analyses. During the post-hoc
analyses, ANOVA tests were performed for comparison of resistance torques
between the three groups, and independent t tests were used for comparison between
two groups. Maximal resistance torque values and catch angles at each actuated
angular velocities were also compared between the three spasticity grades using the
ANOVA test. Catch angles between intermediate and high spasticity grades were
compared using the independent t test.
25
2.2.2 Upper Limb Motion Characterization in Major Movements & Tasks
A. Upper extremity motion capture system and its validation
For upper extremity motion capture, Perception Neuron® (Noitom Ltd., Beijing,
China), a wearable multi-IMU based modular motion capture system was used. In
this study, we utilized 25 IMU sensors for upper body assessment; 3 sensors for body
axis, 4 sensors for each arm, and 7 sensors for each hand including fingers (Fig. 5). A
user interface software, Axis Neuron (Noitom Ltd., Beijing, China), was used for
motion recording and also data extraction. Data sampling rate was set to 60 Hz.
To validate the system’s accuracy and consistency, root mean square error (RMSE)
analyses for elbow flexion / extension and wrist dorsiflexion / volarflexion axis were
performed with electro-goniometer as a reference. In real motion with the system
worn to the body, it is not possible to isolate single joint movement in a single plane
with all other joint fixed. Therefore, coefficient of variation (CoV) analyses for
forearm supination / pronation and elbow flexion / extension for the angles from
gyrosensor, and z-axis (up-down direction) and y-axis (front-back direction) distances
from accelerometers in forearm and hand sensors were performed with the data
collected during the tasks.
26
Fig. 5. (A) A volunteer subject is wearing the IMU-based upper extremity motion capture system. (B) The subject is performing a
task in Action Research Arm Test.
27
B. Participants
Ten healthy volunteers (6 males, 4 females) and nine patients with hemiplegic stroke
were recruited for this study and participated after providing written informed consent.
Their mean age was 29.3 ± 4.7 years old (age range: 23 - 35). Enrolled stroke patients’
mean age was 57.4 ± 17.2 years old (age range: 22 - 73). Four patients had left
hemiplegia and five patients had right hemiplegia. There were 2 (B-stage 3), 2 (B-
stage 4), 3 (B-stage 5), and 2 (B-stage 6) for each Brunnstrom stage (Supplement 2).
Their mean ARAT score was 34.8 ± 21.6 points, where 57 points is the maximum
score.
C. Tasks and procedure
All subjects wore the IMU sensor based motion capture system on both upper
extremities. After sensor calibration, they performed all 19 test items of the ARAT
with both right and left hands alternatively (Supplement 3). They also performed 6
pre-specified ADL tasks: 1) opening a water bottle and drinking, 2) peeling off a
banana, 3) putting on and off of the buttons on a shirt, 4) hair combing, 5) squeezing
toothpaste and toothbrushing, and 6) opening the door knob. These pre-specified ADL
tasks were selected from the survey results from the previous section regarding stroke
patients’ practical needs due to their hemiplegia. During the ADL tasks, the subjects
were instructed to perform the task most naturally, not specifying which hand to hold
or manipulate the object. For all stroke patients, ARAT score and Brunnstrom stage
28
were evaluated.
D. Extracted parameters
Using the Axis Neuron software, acceleration and position data of the wrist and hand
sensors from the accelerometer and Euler angles for sensors of all major joints with
reference to their proximal segment sensors during the ARAT and ADL tasks were
extracted. For each ARAT domain and ADL tasks, the size of workspace in three
orthogonal coordinates and angular position and ROM for each upper extremity joint
were calculated.
For healthy subjects, grasping / pinching and reaching movement during tasks in
ARAT domain 1 and 3 were additionally analyzed regarding initial grasping /
pinching position and ROM during reaching movement.
For all subjects, the average amplitude and maximum amplitude of the movement
segments, and logsum and logsum per time (logsum / time) were extracted and
analyzed. Logsum was defined as the integration of all displacements or changes for
corresponding measurements.
E. Statistical analysis
For validation purposes, intra-subject covariance and inter-subject covariance were
calculated for repetitive grasping / pinching and reaching tasks. Paired t tests were
29
performed to compare workspace dimensions and ROM between dominant and non-
dominant arms. Paired t tests were also performed for comparison of major joint
angles in grasping / pinching position and reaching position, initial position between
grasping and pinching, and reaching position from grasping and pinching. For all
calculated parameters, independent t tests were performed between healthy subjects
and stroke patients. For all parameters that showed significant difference between
healthy subjects and stroke patients, correlation analyses were performed with the
ARAT score as the dependent variable. A p value less than 0.05 was considered
statistically significant.
30
2.3 Feasibility Study for User-intent Driven Robot Control Methods
2.3.1 Image-processing Based Control and its Feasibility
2.3.1.1 Development of an Image-processing Based Hand Rehabilitation Robot
A. Design of a simple two-axis hand rehabilitation robot
In this development, we selected forearm supination / pronation as the essential joint
motion in recovery from stroke. For the execution of the task, a hand grasp / release
motion was included in the design. Based on the fact that most stroke patients
experience proximal limb recovery in the early stage, we assumed that most of the
potential users for this device would have a certain extent of shoulder power and
movement. The forearm support structure was manufactured in the form of a
skateboard with four small wheels mounted at the bottom, so that the user could roll
the whole device freely in any direction with their residual and / or recovered shoulder
movement. Contrary to our initial design, in this pilot study the height adjustment
function was excluded for simplicity. This design structure was intended to make the
device feasible to use at the hospital bedside or at home. The initial design is shown
in Fig. 6.
In the current design, the exoskeleton body part for the hand was placed at the volar
side of the hand, while most of the hand robots’ exoskeleton body are placed at the
dorsum side. This was to prevent hand injury that may be caused by the excessive
grasp motion of the robot.
31
Fig. 6. Initial design for image-processing based user-intent driven 2-axis hand rehabilitation robot is shown.
32
B. Image-processing algorithm for recognizing user-intent
In this hand rehabilitation robot, the camera is mounted on the exoskeleton. This is
different from other types of image guided robot control devices in which most of the
cameras are placed external to the robot in a fixed coordinate system. This concept
comes from a snake-eye view, in contrast to human-eye view. In this robot, the
“image-guided” concept refers to targeting the object and deriving the appropriate
orientation for grabbing the object.
When the patient attempts to grab a target object, the “measurement” button is pressed.
Then the camera detects the long axis of the object, and the software calculates the
difference between the grasp axis and the long axis of the object. Then it
automatically rotates to provide appropriate orientation to grab the object. Once the
robot is oriented in the right position, the user confirms the position and the robot
hand grabs the object. Fig. 7 shows the basic concept and process of the robot.
33
Fig. 7. Main concept of the image-processing based user-intent driven hand
rehabilitation robot is shown. (A) The user aims the robot hand at the target (assumed
as a water bottle with a straw) and confirms the target object shown in the display. (B)
The robot recognizes the target object and the long axis, and automatically rotates the
forearm supination / pronation axis to the appropriate orientation of the robot hand.
(C) The user moves the robot to the object with the proximal muscle power (mainly
shoulder). (D) The user rotates the axis with the controller so that he can drink water.
34
2.3.1.2 Preliminary Usability Test
A. Participants
For the usability test, 20 volunteers were recruited. The participants consisted of six
physicians, five engineers, five rehabilitation therapists, two chronic stroke patients,
and two caregivers of stroke patients. Patients and caregivers were both categorized
as “patients”. All participants were instructed to evaluate the device in their
perspectives of professional experience.
B. Procedure
All participants mounted their left arm and hand on the robot. They were instructed
to use the hand rehabilitation robot freely with the software, which the mounted
camera recognizes the long axis of the target object and automatically rotates to the
appropriate orientation. The users also operate the robot using the switch control pad
with their right hand. After 10 minutes of use, they were asked to fill out the survey
form. The survey consisted of 10 items in a seven-point scale, including sub-items,
asking for the respondent’s overall satisfaction, interest, motivation, expected
improvement in recovery, difficulty, discomfort, safety, comparison to other
therapeutic robots, willingness to use, and expected efficacy after commercialization.
Additional opinions on the robot were also obtained.
35
C. Statistical Analysis
Descriptive statistical analyses were performed.
2.4 Statistical Analysis and Study Approval
All statistical analyses in this study were performed using SPSS v21.0 (SPSS Inc.,
Chicago, IL, USA). All parts of this study involving ethical issues or procedural
justification were approved by the Institutional Review Board of Seoul National
University Hospital (IRB No. 1505-017-668, 1504-104-666, 1610-043-797) and
Seoul National University Boramae Medical Center (IRB No. 20150514/16-2015-
56/061).
36
3. RESULTS
3.1 Practical Robot Functions in Demand
A. Demographic data
A total of 48 subjects (42 men, 6 women) participated in the survey. The mean age
was 42.6 ± 22.2 years and the mean duration since onset was 99.7 ± 54.7 months. All
subjects in unilateral impairment group (UIG, n = 24) had chronic stroke and no
functional use of the hemiplegic arm whereas there were no significant impairments
of the contralateral arm. Bilateral impairment group (BIG, n = 24) consisted of
patients with cervical spinal cord injury (n = 5), Duchenne muscular dystrophy (n =
18), and amyotrophic lateral sclerosis (n = 1). Half of the subjects (n = 12) reported
partial functional use of at least one upper extremity, and the other half (n = 12)
reported no functional use of both upper extremities. Detailed demographic data of
the subgroups are shown in Table 1.
37
Table 1. Demographic Data of Demand Survey Participants
Unilateral impairment
group
Bilateral impairment
group
Total
No. of subjects 24 24 48
Chronic stroke 24
Cervical spinal cord injury 5
Duchenne muscular dystrophy 18
Amyotrophic lateral sclerosis 1
Men : Women (n) 20:4 22:2 42:6
Average age (years) 61.2 ± 5.3 24.8 ± 17.0 42.6 ± 22.2
Average duration since onset
(months)
109.5 ± 51.8 90.1 ± 56.7 99.7 ± 54.6
38
B. Unilateral impairment group (UIG)
In subjects with unilateral upper extremity impairment, handling foods (2.6 ± 1.3,
level of dependence), computer (2.9 ± 1.5), cleaning (3.4 ± 1.5), self-exercise (3.4 ±
1.7) showed high dependency.
Regarding external robotic arm, handling foods (75.0%, important or very important),
using computer (75.0%), hairdressing (70.8%), and using smartphones (66.7%) were
considered objectively important functions as well as subjectively necessary.
However, moving close items (75.0%, necessary or highly necessary) and dressing
(62.5%) were the most highly necessary functions whereas their importance was
relatively rated in lower priority. Assistance in brushing teeth (37.5%, not necessary
or of little necessity), switch control (37.5%), phone calls (33.3%), and eating (33.3%)
showed low necessity with external robotic arms.
Importance and necessity showed generally higher ratings regarding upper limb
exoskeleton. Handling foods (87.5%, important or very important), self-exercise
(87.5%), transfer (87.5%), and smartphone (83.3%) were considered objectively
important functions. Regarding necessity, handling foods (75.0%, necessary or highly
necessary), self-exercise (66.7%), moving close items (75.0%), dressing (66.7%),
washing face (66.7%), hairdressing (66.7%), and smartphone (66.7%) were rated
high. Brushing teeth (66.7%) and eating (62.5%) were relatively rated high in
necessity compared to low rank in importance. Switch control (37.5%, not necessary
or of little necessity), phone calls (37.5%), and wheelchair control (33.3%) were
39
functions that were considered not necessary for an exoskeleton. Detailed data for
UIG are shown in Table 2, Fig. 8, and Fig. 9.
40
Table 2. Dependency, Importance, and Necessity for Assistive Robots in Unilateral Impairment Group (n=24)
ADL item* Dependencya External Robotic Arm Upper Limb Exoskeleton Importanceb Necessaryc Highly
necessaryd Not
necessarye Importanceb Necessaryc Highly
necessaryd Not
necessarye Handling foods 2.6 ± 1.3 75.0 (1) 70.8 (2) 33.3 (4) 20.8 (11) 87.5 (1) 75.0 (1) 45.8 (3) 16.7 (13)
Computer 2.9 ± 1.5 75.0 (1) 54.2 (6) 33.3 (4) 25.0 (7) 70.8 (12) 50.0 (15) 25.0 (14) 25.0 (6) Cleaning 3.4 ± 1.5 45.8 (11) 41.7 (11) 16.7 (10) 20.8 (11) 70.8 (12) 58.3 (12) 29.2 (12) 29.2 (4) Self-exercise 3.4 ± 1.7 87.5 (1) 66.7 (3) 50 (2) 8.3 (18) Moving close items 3.5 ± 1.4 62.5 (6) 75.0 (1) 37.5 (2) 25.0 (7) 79.2 (7) 75.0 (1) 33.3 (11) 20.8 (11)
Dressing 3.5 ± 1.3 54.2 (9) 62.5 (3) 41.6 (1) 16.7 (14) 79.2 (7) 66.7 (3) 54.2 (1) 12.5 (16) Washing Face 3.8 ± 1.6 45.8 (11) 45.8 (9) 25 (7) 29.2 (5) 83.3 (4) 66.7 (3) 41.7 (7) 12.5 (16) Transfer 3.8 ± 1.2 87.5 (1) 58.3 (12) 25 (14) 25.0 (6) Hairdressing 3.8 ± 1.4 70.8 (3) 58.3 (4) 37.5 (2) 20.8 (11) 79.2 (7) 66.7 (3) 45.8 (3) 16.7 (13) Brushing teeth 3.8 ± 1.1 58.3 (7) 33.3 (14) 25 (7) 37.5 (1) 70.8 (12) 66.7 (3) 41.7 (7) 16.7 (13) Phone calls 3.8 ± 1.6 58.3 (7) 37.5 (13) 16.7 (10) 33.3 (3) 79.2 (7) 37.5 (18) 16.7 (17) 37.5 (1) Writing 3.8 ± 1.2 66.7 (4) 54.2 (6) 16.7 (10) 29.2 (5) 75.0 (11) 62.5 (9) 45.8 (3) 25.0 (6) Wheelchair control 3.9 ± 1.3 66.7 (16) 50.0 (15) 29.2 (12) 33.3 (3)
Purse 4.0 ± 1.4 54.2 (9) 50.0 (8) 25.0 (7) 25.0 (7) 66.7 (16) 54.2 (14) 25.0 (14) 25.0 (6) Smartphone 4.0 ± 1.3 66.7 (4) 58.3 (4) 29.2 (6) 25.0 (7) 83.3 (4) 66.7 (3) 37.5 (10) 20.8 (11) Switch control 4.1 ± 1.4 37.5 (14) 41.7 (11) 16.7 (10) 37.5 (1) 58.3 (18) 45.8 (17) 16.7 (17) 37.5 (1) Eating 4.1 ± 1.2 45.8 (11) 45.8 (9) 16.7 (10) 33.3 (3) 70.8 (12) 62.5 (9) 41.7 (7) 29.2 (4) Toilet use 4.3 ± 1.2 83.3 (4) 62.5 (9) 45.8 (3) 25.0 (6)
* ADL items are listed in the order of highest dependence to lowest dependency a dependency in 5-Likert scale: 1(totally dependent) to 5(independent), mean ± standard deviation b percentage of respondents who replied corresponding ADL as important (4) or very important (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group c percentage of respondents who replied corresponding ADL as necessary (4) or highly necessary (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group d percentage of respondents who replied corresponding ADL as highly necessary (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group e percentage of respondents who replied corresponding ADL as not necessary (1) or of little necessity (2) in a 5-Likert scale, numbers in ( ) indicate ranking in the group
41
Fig. 8. The percentage of respondents who replied ‘important’ or ‘very important’ for
importance (dotted line) and ‘necessary’ and ‘highly necessary’ for necessity (solid
line) of an external robotic arm to each activities of daily living (ADL) function in
each impairment group (unilateral impairment group: circle, bilateral impairment
group: rhombus). The ADLs in the x-axis are in the order of high to low dependency
for all survey participants.
42
Fig. 9. The percentage of respondents who replied ‘important’ or ‘very important’ for
importance (dotted line) and ‘necessary’ and ‘highly necessary’ for necessity (solid
line) of an upper limb exoskeleton to each activities of daily living (ADL) function
in each impairment group (unilateral impairment group: circle, bilateral impairment
group: rhombus). The ADLs in the x-axis are in the order of high to low dependency
for all survey participants.
43
C. Bilateral impairment group (BIG)
In subjects with bilateral upper extremity impairment, toilet use (1.6 ± 1.2),
hairdressing (1.7 ± 1.2), dressing (1.7 ± 1.3), transfer (1.8 ± 1.4) and handling foods
(2.0 ± 1.4) showed high dependency.
For an external robotic arm, assistance in eating (70.8%, percentage of being rated
important or very important), dressing (58.3%), phone calls (58.3%), handling foods
(54.2%), and cleaning (54.2%) were considered important functions as well as
necessary functions. Eating (33.3%, percentage of being rated very necessary) and
hairdressing (33.3%) showed highest percentage of being rated highly necessary
functions for an external robotic arm. In contrast, smartphone (58.3%, not necessary
or of little necessity), purse (54.2%), writing (50.0%), and computer (50.0%) showed
low necessity with external robotic arms.
Importance and necessity regarding upper limb exoskeleton also showed generally
higher ratings in the BIG. Wheelchair control (87.5%, 66.7%, important or very
important, necessary or highly necessary, respectively), transfer (83.3%, 66.7%),
toilet use (83.3%, 66.7%), eating (83.3%, 62.5%), phone calls (83.3%, 62.5%),
moving close items (75.0%, 66.7%), dressing (70.8%, 70.8%), handling foods (70.8%,
66.7%), and self-exercise (79.2%, 62.5%) showed both high importance and necessity.
High proportion of subjects replied writing (54.2%) as highly necessary function for
upper limb exoskeleton. Smartphone (41.7%, not necessary or of little necessity),
washing face (37.5%), purse (33.3%), and computer (33.3%) were considered less
44
necessary functions for an exoskeleton. Detailed data for BIG are shown in Table 3,
Fig. 8, and Fig. 9.
45
Table 3. Dependency, Importance, and Necessity for Assistive Robots in Bilateral Impairment Group (n=24)
ADL item* Dependencya External Robotic Arm Upper Limb Exoskeleton Importanceb Necessaryc Highly
necessaryd Not
necessarye Importanceb Necessaryc Highly
necessaryd Not
necessarye Toilet use 1.6 ± 1.2 83.3 (2) 66.7 (2) 45.8 (4) 16.7 (12) Hairdressing 1.7 ± 1.2 50.0 (6) 45.8 (5) 33.3 (1) 16.7 (14) 62.5 (16) 54.2 (13) 33.3 (12) 16.7 (12) Dressing 1.7 ± 1.3 58.3 (2) 50.0 (2) 29.2 (3) 29.2 (11) 70.8 (9) 70.8 (1) 45.8 (4) 12.5 (14) Transfer 1.8 ± 1.4 83.3 (2) 66.7 (2) 58.3 (1) 25.0 (8) Handling foods
2.0 ± 1.4 54.2 (4) 50.0 (2) 25.0 (5) 29.2 (11) 70.8 (9) 66.7 (2) 37.5 (10) 25.0 (8)
Washing Face 2.0 ± 1.6 45.8 (9) 41.7 (6) 12.5 (13) 33.3 (8) 66.7 (14) 50.0 (16) 25.0 (18) 37.5 (2) Self-exercise 2.1 ± 1.6 79.2 (6) 62.5 (7) 41.7 (7) 8.3 (18) Purse 2.2 ± 1.5 37.5 (13) 20.8 (13) 16.7 (10) 54.2 (2) 70.8 (9) 41.7 (18) 33.3 (12) 33.3 (3) Cleaning 2.3 ± 1.2 54.2 (4) 58.3 (1) 29.2 (3) 33.3 (8) 66.7 (14) 62.5 (7) 29.2 (16) 12.5 (14) Switch control 2.3 ± 1.7 41.7 (12) 41.7 (6) 25.0 (5) 41.7 (6) 62.5 (16) 54.2 (13) 29.2 (16) 20.8 (10) Moving close items
2.4 ± 1.3 45.8 (9) 50.0 (2) 12.5 (13) 20.8 (13) 75.0 (7) 66.7 (2) 33.3 (12) 12.5 (14)
Brushing teeth 2.6 ± 1.8 50.0 (6) 37.5 (10) 16.7 (10) 33.3 (8) 70.8 (9) 62.5 (7) 33.3 (12) 29.2 (5) Eating 2.8 ± 1.8 70.8 (1) 41.7 (6) 33.3 (1) 45.8 (5) 83.3 (2) 62.5 (7) 41.7 (7) 29.2 (5) Wheelchair control
3.0 ± 1.4 87.5 (1) 66.7 (2) 58.3 (1) 12.5 (14)
Phone calls 3.1 ± 1.7 58.3 (2) 41.7 (6) 25.0 (5) 41.7 (6) 83.3 (2) 62.5 (7) 45.8 (4) 20.8 (10) Writing 3.1 ± 1.7 45.8 (9) 33.3 (11) 16.7 (10) 50.0 (3) 75.0 (7) 58.3 (12) 54.2 (3) 29.2 (5) Smartphone 3.4 ± 1.7 37.5 (13) 20.8 (13) 20.8 (8) 58.3 (1) 62.5 (16) 45.8 (17) 37.5 (10) 41.7 (1) Computer 3.5 ± 1.7 50.0 (6) 29.2 (12) 20.8 (8) 50.0 (3) 70.8 (9) 54.2 (13) 41.7 (7) 33.3 (3)
* ADL items are listed in the order of highest dependence to lowest dependency a dependency in 5-Likert scale: 1(totally dependent) to 5(independent), mean ± standard deviation b percentage of respondents who replied corresponding ADL as important (4) or very important (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group c percentage of respondents who replied corresponding ADL as necessary (4) or highly necessary (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group d percentage of respondents who replied corresponding ADL as highly necessary (5) in a 5-Likert scale, numbers in ( ) indicate ranking in the group e percentage of respondents who replied corresponding ADL as not necessary (1) or of little necessity (2) in a 5-Likert scale, numbers in ( ) indicate ranking in the group
46
D. Unilateral impaired group vs bilateral impaired group
For external robotic arm, only necessity for smartphone showed significant difference
between unilateral group (3.4 ± 1.5) and bilateral group (2.4 ± 1.6, p = 0.008). For
upper limb exoskeleton, only necessity for wheelchair control showed significant
difference (3.1 ± 1.7 vs 4.1 ± 1.3, unilateral vs bilateral, p = 0.041). Detailed data for
all necessity ratings are shown in Fig. 10.
E. External robotic arm vs upper limb exoskeleton
In most of the cases, upper limb exoskeletons demonstrated higher ratings in
necessity compared to external robotic arm. In UIG, upper limb exoskeleton showed
significantly higher necessity in washing face (3.9 ± 1.2 vs 3.2 ± 1.5, p = 0.004),
brushing teeth (3.8 ± 1.4 vs 3.0 ± 1.5, p = 0.005), and eating (3.5 ± 1.7 vs 3.0 ± 1.5,
p = 0.005). In BIG, significantly higher necessity for upper limb exoskeleton was
shown in dressing (4.0 ± 1.1 vs 3.4 ± 1.4, p = 0.010), moving close items (3.8 ± 1.1
vs 3.4 ± 1.1, p = 0.046), purse (3.2 ± 1.6 vs 2.5 ± 1.4, p = 0.038), switch control (3.5
± 1.3 vs 3.1 ± 1.5, p = 0.038), computer (3.4 ± 1.7 vs 2.7 ± 1.5, p = 0.005), eating (3.5
± 1.7 vs 3.0 ± 1.7, p = 0.020), phone calls (3.7 ± 1.6 vs 3.0 ± 1.6, p = 0.026), writing
(3.7 ± 1.6 vs 2.8 ± 1.5, p = 0.003), and smartphone (3.1 ± 1.8 vs 2.4 ± 1.6, p = 0.008).
Detailed data for all necessity ratings are shown in Fig. 10.
47
Fig. 10. Potential user needs (necessity in 5-Likert scale) are shown for each activities
of daily living (ADL) item regarding external robotic arm and upper limb exoskeleton
in both unilateral (UIG) and bilateral impairment group (BIG). The ADLs in the x-
axis are in the order of high to low dependency for all survey participants. ADLs with
significantly higher necessity for exoskeleton than external robotic arm in UIG
(asterisks, *) and BIG (rhomboids, ◆) are indicated (p < 0.05). Triangles (▲)
indicate significant difference between UIG and BIG (p < 0.05).
48
3.2 Minimum Requirements for Motor Power in Major Joints to
Overcome Spasticity
A. Demographic Data
Demographic data of the subjects are shown in Table 4. Mean age was 60.6 ± 5.4
years and the mean duration since stroke onset was 9.7 ± 3.7 years. Age, height, body
weight, duration since onset, and possible ROM range did not significantly differ
between the spasticity grades, except for wrist flexor spasticity, for which the ROM
range was approximately 16 degrees smaller at the high grade compared to the low
grade of spasticity.
49
Table 4. Demographic Data of Enrolled Stroke Patients for Spasticity Resistance Evaluation
Elbow Flexor Spasticity Wrist Flexor Spasticity Total
Low Intermediate High p* Low Intermediate High p*
Subjects (n) 7 6 7 8 6 6 20
Gender (M/F) 5/2 5/1 7/0 5/3 6/0 17/3
Age (years old) 60.6±5.5 57.5±7.4 62.1±4.2 0.315 61.5±5.9 59.2±4.8 60.8±6.0 0.592 60.6±5.4
Height (cm) 161.8±9.0 171.8±4.0 166.5±4.3 0.092 162.1±9.3 167.7±3.9 168.3±5.9 0.345 165.7±7.3
Body weight (kg) 66.3±9.4 73.3±12.4 71.6±9.3 0.593 66.0±9.1 73.8±9.7 70.8±10.8 0.408 69.8±9.9
Etiology (Inf / Hem / Etc)
0/6/1 3/3/0 4/3/0 0/7/1 3/3/0 4/2/0 7/12/1
Laterality (left / right)
5/2 2/4 3/4 5/3 3/3 2/4 10/10
Years Since Onset 9.8±4.3 8.8±3.6 10.1±3.4 0.560 9.9±4.3 9.2±2.9 10.0±4.0 0.867 9.7±3.7
Range of Motion (°) 110.0±23.3 112.5±9.6 106.9±18.3 0.887 132.5±17.9 126.8±15.1 116.5±24.4 <0.001 Elbow: 109.3±18.5 Wrist: 126.1±20.3
Low: MAS grade 0, 1 ; Intermediate: MAS grade 1+ ; High: MAS grade 2, 3 ; Inf: infarction ; Hem: hemorrhage ; Etc: others, e.g. brain tumor
* Kruskal-Wallis test between spasticity grades, p < 0.05 considered significant
50
B. Torque Resistance Response for Elbow Flexors
The maximal resistance torques caused by the elbow flexor with low, intermediate,
and high grade spasticity averaged over all angular velocities were 3.68 ± 2.42, 5.94
± 2.55, and 8.25 ± 3.35 Nm, respectively, with statistically significant differences
between the grades (p < 0.001 by ANOVA and all post-hoc analyses). The maximum
resistance torque for the elbow flexor was 1.77 Nm in a healthy subject. The
maximum resistance torque value during extension of the elbow joint among all
subjects and trials was 21.28 Nm, which occurred in the subject with MAS grade 2
in the elbow flexor. The RM-ANOVA test for torque values at 1/3, 2/3, and 100 %
ROM demonstrated that the resistance torque value showed significant differences
between spasticity grades showing a tendency of increase with increasing spasticity
(p < 0.001 for interaction of ROM and spasticity grade by RM-ANOVA). The
ANOVA tests for resistance torques at 1/3 ROM, 2/3 ROM, and 100 % ROM also
demonstrated significant differences between spasticity grades showing increasing
resistance torque with increasing spasticity (p = 0.001 for 1/3 ROM, p < 0.001 for 2/3
ROM and 100 % ROM). The results for elbow flexors are shown in Table 5 and Fig.
11.
C. Torque Resistance Response for Elbow Extensors
For the elbow extensor, the maximal resistance torques averaged over all angular
velocities were -4.95 ± 3.30, -9.65 ± 4.35, and -8.35 ± 3.66 Nm for low, intermediate,
and high grade spasticity, respectively (p < 0.001 by ANOVA, p < 0.001 for low vs
intermediate, low vs high grade spasticity), compared to -2.82 Nm in a healthy subject.
51
The maximum resistance torque during flexion of the elbow joint among all subjects
and trials was -18.84 Nm, which occurred in a subject with MAS grade 1+ in both
the elbow flexor and extensor. The RM-ANOVA test for resistance torques at 1/3,
2/3, and 100% ROM showed significant interaction of ROM and spasticity grades (p
< 0.001). The ANOVA test regarding resistance torques for 1/3, 2/3, and 100% ROM
and maximal resistance torque showed significant differences between the spasticity
grades (p < 0.01); however, post hoc analysis demonstrated that the resistance torque
at 100% ROM and the maximal value were not significantly different between
intermediate and high spasticity. The results for elbow extensors are shown in Table
5 and Fig. 11.
D. Torque Resistance Response for Wrist Flexors
The maximal resistance torques caused by the wrist flexor averaged over all angular
velocities were 4.23 ± 1.75, 5.68 ± 1.96, and 5.44 ± 2.02 Nm for low, intermediate,
and high grade spasticity, respectively (p < 0.001 by ANOVA, p < 0.001 for low vs
intermediate, low vs high grade spasticity), compared to 0.56 Nm in a healthy subject.
The maximum resistance torque during extension of the wrist joint among all subjects
and trials was 11.34 Nm, which occurred in a subject with MAS grade 2 in the wrist
flexor. The maximum resistance torque in the flexion direction was -7.5 Nm among
all subjects, who did not show specific wrist extensor spasticity. The RM-ANOVA
test for resistance torques at 1/3, 2/3, and 100% ROM showed significant interaction
of ROM and spasticity grades (p < 0.001). The ANOVA test for torque values at 1/3,
2/3, and 100% ROM also showed significant differences between spasticity grades
52
(p < 0.001), while resistance torque at 100% ROM between intermediate and high
spasticity grades was not significantly different (p = 0.869). Because intermediate
and high spasticity did not show meaningful difference, the independent t test was
performed as a part of post-hoc analysis between low grade and intermediate plus
high grades. The resistance torques at 2/3, 100% ROM and the maximal value were
significantly different (p < 0.001). The results for wrist flexors are shown in Table 5
and Fig. 11.
53
Table 5. Resistance Torque Values to Spastic Elbow and Wrist Joints
Spastic Muscle (No. of trials for low, intermediate, high grade spasticity)
Angular Position
Reference (healthy subject)
Low Grade Spasticity (MAS 0,1)
Intermediate Grade Spasticity (MAS 1+)
High Grade Spasticity (MAS 2,3)
p value (ANOVA)
Posthoc Analysis Low vs Intermediate
Low vs High
Intermediate vs High
Elbow Flexor (75/36/78)
0%ROM -2.17 -3.61±2.52 -4.06±2.62 -4.36±2.05 0.150 >0.05 >0.05 >0.05 1/3ROM 0.99 0.06±0.84 0.83±1.57 0.93±1.81 0.001 0.025 0.001 0.939 2/3ROM 0.79 1.66±1.80 4.05±2.99 4.64±2.77 <0.001 <0.001 <0.001 0.463 100%ROM 0.37 3.44±2.45 5.57±2.05 7.85±3.34 <0.001 0.001 <0.001 <0.001 Maximal Torque 1.77 3.68±2.42 5.94±2.55 8.25±3.35 <0.001 <0.001 <0.001 <0.001
Elbow Extensor (138/17/34)
0%ROM 0.70 4.04±2.13 2.66±1.31 4.37±1.34 0.011 0.018 0.653 0.010 1/3ROM 0.79 0.39±0.83 -0.27±0.74 0.56±0.82 0.003 0.006 0.501 0.002 2/3ROM 1.16 -1.64±2.03 -4.72±3.29 -2.40±1.76 <0.001 <0.001 0.157 0.001 100%ROM -2.68 -4.86±3.15 -9.31±4.49 -7.79±3.42 <0.001 <0.001 <0.001 0.281 Maximal Torque
-2.82 -4.95±3.30 -9.65±4.35 -8.35±3.66 <0.001 <0.001 <0.001 0.418
Wrist Flexor (96/71/68)
0%ROM -0.19 -1.30±1.19 -1.58±1.28 -2.03±1.77 0.006 0.412 0.004 0.154 1/3ROM -0.08 0.05±0.66 0.56±0.80 -0.22±0.92 <0.001 <0.001 0.079 <0.001 2/3ROM -0.07 1.17±0.87 2.07±1.26 1.54±1.34 <0.001 <0.001 0.096 0.021 100%ROM 0.44 4.14±1.75 5.52±1.93 5.36±2.03 <0.001 <0.001 <0.001 0.869 Maximal Torque
0.56 4.23±1.75 5.68±1.96 5.44±2.02 <0.001 <0.001 <0.001 0.730
MAS: modified Ashworth Scale; ANOVA: analysis of variance; ROM: range of motion
54
Fig. 11. Box-and-whisker plots are shown for maximal resistance torque of (A) elbow flexor, (B) elbow extensor, and (C) wrist flexor.
Analysis of variance test showed significant differences between spasticity grades. Asterisks indicate significant differences between
the grades by posthoc analysis (p < 0.05).
55
E. Effect of Robot Speed
In each spasticity grade for elbow flexors, the maximal resistance torques were
compared by the angular velocity of the isokinetic robotic movement. There were no
significant differences according to the angular velocities (p = 0.231, 0.478, 0.766),
but a trend was shown for the resistance torque to increase with velocity. In contrast,
the CPP demonstrated a significant tendency to decrease with increasing angular
velocity in subjects with low and intermediate spasticity, which means that the catch
occurred more quickly when the robotic movement speed was faster (p = 0.009 and
0.004). Data are shown in Fig. 12.
For wrist flexor spasticity, the robot speed did not significantly affect the maximal
resistance torque, nor did it show any specific tendency relative to angular velocity
(p = 0.917, 0.697, and 0.989 for low, intermediate, and high spasticity, by ANOVA,
data not shown).
F. Catch Point Percentage
The catch angle for each trial was converted into CPP (%ROM) and the average CPP
was calculated by the degree of spasticity. The estimated CPP for intermediate and
high elbow flexor spasticity was 57.56 ± 30.91 %ROM and 49.11 ± 21.32 %,
respectively (p < 0.001). The mean torque at the moment of the catch did not show
significant differences between the spasticity grades (data not shown). With regard to
the wrist flexor spasticity, the CPP for intermediate and high spasticity were 73.2 ±
22.4 and 76.3 ± 23.5 %ROM, respectively (p > 0.05).
56
Fig. 12. (A) Maximum resistance torque and (B) catch angle according to the angular velocity of the robot in each grade of elbow
flexor spasticity are shown. Catch angle for low (MAS 0, 1) and intermediate (MAS 1+) spasticity showed significant decrease
with increasing velocity.
57
3.3 Range of Motion and Movement Characteristics in Major
Movements & Tasks
3.3.1 Healthy Subjects
A. Validation of the upper extremity motion capture system
Validation was performed for 6 reaching tasks within the ARAT tasks for 10 healthy
subjects. The range of RMSE for elbow flexion / extension angle ranged from 2.11°
to 4.75° (average: 3.61 ± 1.32°), and 0.42° to 1.22° (average: 0.85 ± 0.40°) for wrist
dorsiflexion / volarflexion angle. During the reaching task, the mean change of
forearm supination / pronation was 36.65 ± 6.98°, with intra-subject CoV of 17.29%
and inter-subject CoV of 19.05%. The change of elbow flexion / extension was 69.96
± 16.89°, and intra-subject and inter-subject CoV were 11.67% and 24.14%,
respectively. Distance data extracted from the sensors during the reaching tasks were
evaluated and also compared with real movement distance. Regarding the
accelerometer on the forearm sensor, average of the calculated movement distance
were 34.14 ± 4.15 cm in z-axis and 33.54 ± 4.79 cm in y-axis, where measured
distance in each direction was 34.0 cm and 33.5 cm, respectively. Data calculated
from hand sensors were 36.78 ± 3.09cm and 32.35 ± 4.64 cm, respectively. Intra-
subject CoV ranged from 5.5 to 9.5% while inter-subject CoV ranged from 8.4 to
14.3%. Full results are shown in Table 6.
58
Table 6. Coefficient of Variation during Major Movements for IMU-based Upper Extremity Motion Capture System
Sensor type Movement / direction Mean±SD of change during task (across subjects)
Intra-subject CoV average
Inter-subject CoV
Estimated real distance
Gyrosensor Forearm supination/ pronation
36.65 ± 6.98° 17.29% 19.05% -
Elbow flexion/ extension
69.96 ± 16.89° 11.67% 24.14% -
Accelerometer (forearm sensor)
z-axis distance* (up/down)
34.14 ± 4.15 cm 6.18% 12.17% 34.0 cm
y-axis distance* (front/back)
33.54 ± 4.79 cm 7.16% 14.28% 33.5 cm
Accelerometer (hand sensor)
z-axis distance* (up/down)
36.78 ± 3.09 cm 5.56% 8.41% 34.0 cm
y-axis distance* (front/back)
32.35 ± 4.64 cm 9.49% 14.33% 33.5 cm
* Estimated distance between initial object position and target position is approximately 39cm for y-axis and 40cm for z-axis. Note that this is distance regarding center of object, while calculated distance from accelerometers refers to position of the sensor (forearm and hand dorsum).
59
B. Workspace and ROM in basic upper extremity movements
All 10 subjects were right-handed. For orthogonal coordination, axes were defined as
following: left-right direction as x-axis, front-right direction as y-axis, and up-down
direction as z-axis. For the ARAT tasks, size of the workspace for the right hand with
reference to the sensor on the dorsum of the hand was 0.53 ± 0.11 m for x-axis, 0.92
± 0.08 m for y-axis, and 0.89 ± 0.10 m for z-axis. For the left side, average workspace
size was 0.62 × 0.80 × 0.86 m (in x, y, z-axis order). For pre-specified ADL tasks, the
workspace for the dominant hand was 0.71 ± 0.22 m, 0.70 ± 0.17 m, and 0.86 ± 0.11
m (in x, y, z-axis order). Workspace of the non-dominant hand was significantly
smaller, with average size of 0.52 × 0.53 × 0.65 m (p = 0.001, 0.011, and 0.001 for x,
y, and z-axis, respectively). Detailed data are shown in Table 7.
For ROM in major upper extremity joints, the angular range was similar between
right and left sides. Elbow flexion / extension and forearm supination / pronation
showed highest value for ROM in both ARAT and ADL for the dominant arm. The
ROM values were 109.15 ± 18.82° and 105.23 ± 15.38° (elbow flexion / extension
and forearm supination / pronation, respectively) for ARAT tasks and 120.61 ± 23.64°
and 128.09 ± 22.04° for ADL tasks. The ROM of the dominant side were significantly
greater than the non-dominant side for all joint directions except wrist dorsiflexion /
volarflexion, which showed similar values (113.70 ± 18.26° vs 110.08 ± 12.16°, right
vs left, p = 0.526). Detailed data are shown in Table 7.
60
Table 7. Range of Motion Angle between Right and Left Upper Extremities During ARAT and
ADL Tasks
Axis Right Left pa ARAT x-axis (left-right, hand sensor) 0.53 ± 0.11m 0.62 ± 0.07m 0.082 y-axis (front-back, hand sensor) 0.92 ± 0.08m 0.80 ± 0.11m 0.049* z-axis (hand sensor) 0.89 ± 0.10m 0.86 ± 0.08m 0.224 Shoulder abduction/adduction 50.16 ± 11.14° 55.34 ± 13.48° 0.249 Shoulder flexion/extension 79.52 ± 19.34° 75.71 ± 21.56° 0.478 Elbow flexion/extension 109.15 ± 18.82° 106.89 ± 12.83° 0.705 Forearm supination/pronation 105.23 ± 15.38° 108.64 ± 12.64° 0.426 Shoulder internal/external rotation 91.99 ± 20.98° 84.44 ± 44.75° 0.584 Wrist dorsiflexion/volarflexion 82.90 ± 22.52° 81.26 ± 11.16° 0.833 ADL tasks
x-axis (left-right, hand sensor) 0.71 ± 0.22m 0.52 ± 0.13m 0.001* y-axis (front-back, hand sensor) 0.70 ± 0.17m 0.53 ± 0.15m 0.011* z-axis (hand sensor) 0.86 ± 0.11m 0.65 ± 0.13m 0.001* Shoulder abduction/adduction 58.84 ± 14.53° 35.43 ± 10.09° <0.001* Shoulder flexion/extension 68.41 ± 17.56° 40.49 ± 18.54° 0.002* Elbow flexion/extension 120.61 ± 23.64° 102.53 ± 19.51° 0.044* Forearm supination/pronation 128.09 ± 22.04° 108.00 ± 16.23° 0.027* Shoulder internal/external rotation 111.56 ± 31.88° 77.04 ± 21.28° 0.030* Wrist dorsiflexion/volarflexion 113.70 ± 18.26° 110.08 ± 12.16° 0.526
a p value for paired t test between right and left side
* p value less than 0.05 considered statistically significant
61
C. Characteristics of grasping / pinching and reaching
Upper extremity posture during grasping / pinching and reaching was analyzed as a
subset analysis of the motion data extracted from grasping / pinching and reaching
tasks in ARAT domains 1 and 3. Comparing grasping and pinching posture, shoulder
was more significantly abducted during pinching (19.39 ± 7.84°) compared to
grasping (15.33 ± 6.91°, p = 0.040) and more extended during pinching (29.12 ±
12.33°) than grasping (22.99 ± 10.63°, p = 0.038). Elbow flexion / extension, forearm
supination / pronation and shoulder internal / external rotation did not significantly
differ between the two postures. During reaching after grasping, elbow was extended
for 87.87 ± 25.18° from initial flexed posture and pronated for 36.65 ± 6.98° from
initial posture. The degree of elbow extension and forearm pronation while reaching
after pinching were similar (p = 0.849 and 0.294, respectively). Detailed results are
shown in Table 8.
62
Table 8. Major Joint Angle Position and Change During Grasping / Pinching and Reaching
Axis Grasping Initial
Position
ROM during Reaching
p Pinching Initial
Position
ROM during Reaching
p Grasp-Pinch
pa
Reaching difference pb
Shoulder abduction/ adduction
15.33±6.91° (abduction)
22.48±19.81° (toward
abduction)
0.006* 19.39±7.84° (abduction)
23.67±13.35° (toward
abduction)
<0.001* 0.040* 0.015*
Shoulder flexion/extension
22.99±10.63° (extension)
47.80±17.70° (toward flexion)
<0.001* 29.12±12.33° (extension)
41.83±13.69° (toward flexion)
<0.001* 0.038* 0.948
Elbow flexion/extension
87.87±25.18° (near fully
flexed)
69.96±16.89° (toward
extension)
<0.001* 84.82±20.25° (near fully
flexed)
67.91±14.16° (toward
extension)
<0.001* 0.543 0.849
Forearm supination/pronationc
34.37±11.07° (supinated)
36.65±6.98° (toward
pronation)
<0.001* 30.98±13.71° (supinated)
36.02±12.44° (toward
pronation)
<0.001* 0.181 0.294
Shoulder IR/ER
0.68±23.56° (inward
direction)
16.55±23.02° (toward external
rotation)
0.049* 2.01±13.74° (inward
direction)
18.10±13.02° (toward external
rotation)
0.002* 0.794 0.860
Wrist deviation 8.94±12.12° (to thumb
side)
-1.76±10.21° (to finger side)
0.599 1.05±8.19° (to thumb
side)
4.81±8.85° (to thumb side)
0.120 0.004* 0.522
Wrist rotation 4.59±7.35° (toward palm
down)
7.12±4.59° (toward palm
up)
0.001* 0.80±5.25° (toward palm
down)
4.75±4.05° (toward palm
up)
0.005* 0.023* 0.385
Wrist dorsiflexion / volarflexion
18.79±16.35° (dorsiflexed)
6.79± 6.00° (toward
volarflexion)
0.006* 11.30±13.90° (dorsiflexed)
7.28±11.22° (toward
volarflexion)
0.070 0.166 0.123
a Comparison between grasping and pinching posture by paired t test b Comparison between ROM change during reaching after grasping and pinching by paired t test c Full pronation: 0°, full supination: 180° * p value less than 0.05 considered statistically significant
63
3.3.2 Stroke Patients
Of the parameters that showed significant differences in values between healthy
subjects and patients and also significant correlation with clinical measures, the
average amplitude of forearm supination / pronation angle during the ARAT domain
4 tasks demonstrated the greatest decline in the value of severely impaired patients
compared to healthy subjects (29.83%) and also the largest difference between
severely and mildly impaired patients (48.46%). During ADL tasks, logsum per time
for supination / pronation showed a profound difference between severity levels
(38.33%). The average amplitude of acceleration in the x-axis (left-right) and z-axis
(up-down) of the hand and wrist sensors during the ARAT tasks demonstrated a range
of 45 to 60% value compared to healthy subjects, with a 21.6 to 37.8 % difference
along the severity spectrum. Detailed results are shown in Table 9 and Fig. 13.
64
Table 9. Average Amplitude Angles and Acceleration for Significantly Declined Parameters in Stroke Patients by Brunnstrom Stage Task Position / Direction Brunnstrom stage Normal Difference between
severely and mildly impaireda
Residual value in severely impairedb 3 4 5 6
Angle (gyrosensor data, degree) ARAT full Wrist
dorsiflexion/volarflexion 7.23 8.62 11.61 14.31 13.57 52.17% 53.28%
ARAT domain 4
Forearm supination/pronation
11.40 16.14 23.06 29.92 38.22 48.46% 29.83%
ARAT domain 1
Forearm supination/pronation
8.67 10.56 14.08 16.78 18.03 44.98% 48.09%
ARAT domain 1
Elbow flexion/extension 9.12 14.49 22.29 22.23 32.91 39.84% 27.71%
ARAT domain 3
Elbow flexion/extension 9.12 15.32 19.23 21.96 32.28 39.78% 28.25%
ARAT domain 3
Shoulder internal/external rotation
7.78 10.70 14.60 13.45 15.60 36.35% 49.87%
ARAT domain 2
Wrist dorsiflexion/volarflexion
7.71 7.73 9.88 12.38 13.68 34.14% 56.36%
ARAT full Forearm supination/pronation
9.94 12.71 14.89 16.15 19.30 32.18% 51.50%
ARAT full Elbow flexion/extension 10.49 14.67 19.33 19.39 28.21 31.55% 37.19% Acceleration (accelerometer data, m/s2) ARAT full Hand x-axis 0.22 0.24 0.31 0.35 0.37 35.14% 59.46% ARAT full Hand z-axis 0.25 0.32 0.35 0.39 0.47 29.79% 53.19% ARAT full Hand y-axis 0.21 0.30 0.33 0.32 0.40 27.50% 52.50% ARAT full Wrist x-axis 0.18 0.20 0.28 0.25 0.29 24.14% 62.07% ARAT full Wrist y-axis 0.17 0.22 0.27 0.25 0.37 21.62% 45.95%
a Percentage value of (B-stage 6 – B-stage 3) / Normal b Percentage value of B-stage 3 / Normal
65
Fig. 13. Average amplitude angles of joint movement segments during ARAT tasks by Brunnstrom stage are shown.
66
3.4 Preliminary Usability Test for Image-processing Based Hand
Rehabilitation Robot
For overall satisfaction regarding the robot’s ability to help stroke rehabilitation,
physiatrists replied with the highest score (6.0 ± 0.9), followed by robot engineers
(5.4 ± 0.5), therapists (4.6 ± 0.5), and patients (4.0 ± 1.2). Participants found the
device interesting (5.7 ± 1.2), motivating (5.8 ± 0.9), and also having less possibility
of injury or safety issues (6.1 ± 1.1). However, the level of difficulty (4.8 ± 1.9),
expectance of improvement (5.1 ± 1.0), and comfort (4.9 ± 1.3) were relatively low.
Detailed response results are shown in Fig. 14.
67
Fig. 14. Usability test results for an image-processing based hand rehabilitation robot are shown for 4 categories of respondents.
The scores are in a 7-point scale with 7 points representing most satisfactory or safe and 0 points for least satisfactory or unsafe.
68
4. DISCUSSION
4.1. Demand survey for potential users of robots
Highly rated ADLs should be taken into consideration when designing a BMI
controlled robot for rehabilitation or ADL assistance, especially if the robot is going
to be simple and portable with selected functions. There have been several previous
studies on the functional priorities of patients with specific diseases [34-36]. However,
to our knowledge, this is the first study to assess the needs of potential users by the
type of robot and laterality of the impairment.
Chronic stroke patients with hemiplegia generally rated bimanual ADLs as the most
necessary functions for both types of robot. Nearly 80% of stroke patients who have
hemiplegia become able to perform most one-handed ADLs and several two-handed
ADLs according to the degree of hemiparesis. All 24 stroke patients in this study also
had hemiplegia and were able to use the unaffected arm properly. Therefore, ADLs
that require both upper extremities were rated as the most necessary functions that a
robot should provide. Handling foods, dressing, and hairdressing are functions that
require both hands for its performance. However, the most important and necessary
function for an exoskeleton in stroke patients was self-exercise of upper extremity,
getting a higher score than bimanual functions. The mean duration since stroke onset
was about 110 months in our study population. Many of these chronic stroke patients
had moderate to severe spasticity in major upper extremity joints. The rate of
69
spasticity after stroke is reported to be approximately 30 to 60 percent [37]. During a
short interview after the survey, participants stated that they would like to have an
upper extremity robot that could help them exercise at home, in addition to the
therapies that they receive in rehabilitation centers. It seems that personal robots with
exercise functions would help patients continue their exercises at home to enhance
functional recovery and decrease spasticity, which would increase their ADL
performance level. It has also been reported in many studies that a decrease in
spasticity leads to better ADL performance and a lower burden for care [14, 38]. In
addition, several studies showed greater functional improvement when robotic
rehabilitation therapy was provided for up to 5 hours a day for 12 weeks compared to
that of a lesser treatment dose, which indicates that continuing exercise with robots
throughout the day would result in better outcomes [39-41].
In BIG, eating and hairdressing as well as cleaning, handling foods, dressing, and
moving close items were ADLs that received relatively high scores for the necessity
of external robotic arms. For exoskeletons, dressing, toilet use, transfer, wheelchair
control, moving close items, and handling foods showed high demand. A previous
study on amyotrophic lateral sclerosis (ALS) patients also showed that “using the
bathroom” was among the highest priorities for a BMI in addition to communication
and controlling motorized wheelchairs [35]. Self-exercise of the upper extremity also
received the high scores in both importance and necessity. Self-exercise function of
the robot in this survey was defined as the ability of the exoskeleton, with or without
any kind of user-robot interaction, to provide passive ROM or active ROM exercises,
70
in the manner that the user may select and control the moving joint and the extent of
ROM. This result is in agreement with previous studies, which found that patients
with spinal cord injury replied that restoration of walking and arm and hand functions
along with bladder and bowel control were all high priorities, and that they would
like to use a BMI to control functional electrical stimulation in order to enhance
functional recovery [34, 42]. Patients in this study wanted to control the exoskeleton
with a BMI to perform their upper extremity exercises.
Fig. 1 and 2 shows the importance and necessity of each type of robot in both
impairment subgroups. The ADLs in the x-axis are given in the order of level of
dependency, from highly dependent to near independent. It is easily noticed that the
importance and necessity is not in proportion with dependency. The results in these
figures show ADLs that were considered important, but not actually necessary for
everyday lives. In regards to both robot types, UIG did not need to do phone calls or
use computers with the robot. This may be because they have an intact arm, and are
older in age compared to patients in other disease categories. For exoskeletons, stroke
patients did not require wheelchair control or transfer assistance, as they could
manage it with their intact arm or using a cane.
In general, the ratings for importance and necessity for the exoskeleton were higher
than those for the external robotic arm as shown in Fig. 3. UIG showed significantly
higher necessity for exoskeleton compared to external robotic arm in washing face,
brushing teeth, and eating. In BIG, dressing, moving close items, purse, switch
control, computer, eating, phone calls, writing, and smartphone showed higher
71
necessity for exoskeletons. While bimanual ADLs such as hairdressing, handling
foods and dressing showed high scores for both types of robot, patients with unilateral
impairment demonstrated higher demands for using exoskeletons for grooming
related activities, whereas subjects with bilateral impairments tended to give higher
scores to activities necessary for social functioning and interactions.
4.2. Biomechanical Response of Exoskeleton to Spasticity
The main results of the spasticity study were resistance torque values for robotic
elbow and wrist joints during actuation for patients with spastic arms. The mean
torque values for various levels of spasticity in the elbow flexor ranged from 3.68 to
8.25 Nm, with statistically significant differences between the grades. Considering
that the maximal resistance torque in a healthy subject was 1.77 Nm and the
calculated torque required for maintaining isokinetic rotation of the forearm and the
hand (not the robot) from the elbow joint against gravity based on the human database
(Size Korea, http://sizekorea.kats.go.kr) is approximately 0.97 Nm and 1.14 Nm,
respectively, the amount of resistance produced by spasticity during the isokinetic
actuation is considerable [43]. Park et al. [44] showed that the resistance torque
measured during physical examination was approximately 3 Nm in MAS grade 1+
and between 4 and 6 Nm in MAS grades 2 and 3 for individual subjects. In their study,
passive movement conducted by an examiner was paused during the moment of the
catch, and was continued after the resistance was decreased. In the present study, the
robotic joint maintained isokinetic movement without giving the spastic joint
72
sufficient time for the resistance to be decreased, resulting in higher resistance values
than those of the previous study. This resistance pattern is consistent with other
studies, in that the resistance torque kept increasing beyond the point of maximal
stiffness with continuation of the movement, as observed in various situations such
as passive movement by an examiner [33], both slow and fast isokinetic movements
[45, 46], and even in active and non-isokinetic cases [47]. In addition, the mean
torques measured at the 2/3 ROM point during extension of the elbow joint were 4.05
and 4.64 Nm, respectively, for intermediate and high spasticity, which is similar to
the results of a previous study [44]. Considering that the maximal value measured
during the study was 21.28 Nm, it appears that maximum output torque range buffer
for the spastic elbow flexor component need not to exceed 21.28 Nm. However, it
does need to be modified according to the target population or desired function of the
robot. The required torque for each joint to actuate the robot would include the torque
needed to rotate the robot frame distal to the joint (τrobot), human arm mass distal to
the joint (τarm), and torque to overcome spasticity (τspasticity), which can be calculated
approximately as shown in Equation 3:
𝑑𝑑𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = τrobot frame + 𝑑𝑑𝑡𝑡𝑎𝑎𝑎𝑎 + 𝑑𝑑𝑠𝑠𝑠𝑠𝑡𝑡𝑠𝑠𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠𝑡𝑡𝑠𝑠
= 𝑟𝑟𝐶𝐶𝐶𝐶𝑀𝑀𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑓𝑓𝑟𝑟𝑓𝑓𝑓𝑓𝑓𝑓 × 𝐹𝐹𝑎𝑎𝑡𝑡𝑟𝑟𝑡𝑡𝑡𝑡 + 𝑟𝑟𝐶𝐶𝐶𝐶𝑀𝑀𝑓𝑓𝑟𝑟𝑓𝑓 × 𝐹𝐹𝑡𝑡𝑎𝑎𝑎𝑎 + 𝑑𝑑𝑠𝑠𝑠𝑠𝑡𝑡𝑠𝑠𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠𝑡𝑡𝑠𝑠 (3)
where COM represents center of mass (COM), r is distance from the joint axis to the
COM, and F is force. Parameters rCOMrobot frame and rCOMarm represents the distance from
the joint axis to the center of mass of the robot frame and human arm, respectively,
and Frobot and Farm refers to the force needed to move the robot frame and human arm
73
to perform intended movement, respectively.
In most chronic stroke patients, spasticity for the elbow joint usually appears in the
elbow flexor muscles; however, in some patients, spasticity exists on the opposite
side, or even on both sides [17]. Seven patients in this study also showed elbow
extensor spasticity. The number of subjects and trials for elbow extensor spasticity
was relatively small, but results suggested that resistance torques for low spasticity
and intermediate to high spasticity have significant differences. Maximal resistance
torques were in a similar range to results for elbow flexor spasticity. The data reported
by Starsky et al. [46] showed a similar range during the flexion of an elbow joint with
a spastic elbow extensor, but detailed data were not provided. It seems that similar
specifications may be applied for the flexion direction of the elbow joint.
For the wrist flexors, the maximal resistance torque values showed statistically
significant differences between low and intermediate to high grades of spasticity, and
the differences of the values between the spasticity levels were relatively small
compared to those for the elbow. However, the resistance created by the wrist flexor
spasticity was nearly 10 times that of the healthy subject. Despite the fact that
spasticity of the wrist flexors has been a critical problem in many stroke patients, it
seems that mechanical assessment of the wrist flexors has been rare because the
resistance torque value is relatively small owing to the fact that it is a distal and
smaller joint. Malhotra et al. [48] evaluated the stiffness of wrist flexors by levels of
spasticity, which showed no significant differences between spasticity levels;
however, the maximal resistance torque values were not presented. Many
neurorehabilitation robots in wearable form with shoulder joints do not have a wrist
74
joint [49, 50], while large robots for use in treatment already possess motors in their
wrist joints that can simply overcome these amounts of resistance [51, 52]. However,
in the design of neurorehabilitation robots for portable use in daily living, wrist flexor
spasticity must be considered.
As spasticity is commonly defined as “velocity dependent” [13], analysis regarding
angular velocity of the robot joint was performed. In each MAS grades of elbow
flexor spasticity, the maximum resistance torque showed a tendency to increase with
increasing velocity; however, statistical analysis did not support this tendency. Seth
et al. [53] reported that the robot’s velocity had a significant effect on robot resistance
for the elbow joint; however, their study was performed on healthy subjects and
patients with MAS grade 0, which may not be applicable to higher spasticity grades.
However, the number of subjects and trials in the present study may have been
insufficient to ensure sufficient statistical power. The CPP presented by %ROM
demonstrated a significant decreasing tendency with higher velocity, which means
that the catch occurred more quickly at higher velocities. This is consistent with
clinical experiences and a study by Wu et al. [54] performed on patients with cerebral
palsy using a manual spasticity evaluator. They found that the catch angle occurred
later with increasing velocity because the examiner manually extended the elbow
more quickly and with greater strength despite the increased resistance and early
activated spastic muscle EMG signal, resulting in a greater catch angle. As mentioned
earlier, it is known that spasticity consists of two major components: reflexive and
tissue components [13]. The reflexive component is strongly associated with the catch
phenomenon, whereas the tissue component is related to stiffness throughout the
75
passive movement of the joint. The velocity-dependent feature of spasticity stems
from the reflexive component, and therefore, the catch angle may be affected by the
angular velocity; however, the maximal resistance appears at the end of the ROM
range and this would not be affected by the velocity, but rather the level of spasticity.
For practical utilization of a portable and wearable neurorehabilitation robot, it is
important that the robot weight is made as light as possible. Potential users with
impaired limbs require robots of even lighter weight than those healthy people may
wear. However, robots must have sufficient power and capabilities as well as an
appropriate control algorithm to help the user perform the desired movement and
activities. Robots should be able to deal with spasticity in an appropriate and safe
manner with a certain amount of output torque. In addition to setting an adequate
torque range for spasticity induced resistance, other methods may be applied to create
a light-weight exoskeleton. One possible method is to apply a pause during movement
when the resistance exceeds the threshold of the motor. At the end of the ROM where
the robot is actuated against spasticity, the activity of the spastic muscle remains for
a certain amount of time with an exponential decay of the resistance torque (Fig. 2).
Using this phenomenon, which is equivalent to the decrease of resistance during the
catch while the examiner is still applying force to the joint, the robot can still be
applied to a spastic limb, even with a low-torque output motor without resulting in
excessive loading. A precise sensing and control system would be necessary for this
type of system.
76
4.3. Kinematic Characteristics of Upper Extremity in Healthy
Subjects and Stroke Patients
The purpose of this study was to provide clinically relevant information regarding
workspace and major joint angle range while performing essential ADLs or important
movements. By identifying these factors, it is possible to limit the extent of
exoskeleton movements and therefore modify the design of the robot so that it can
move within the designated workspace with relatively more simple structure. In this
study we evaluated the ROM and workspace during performing the ARAT tasks, one
of the common functional evaluation tool in the clinics, because it is well known to
significantly correlate with the patients’ functional status or recovery state [55-57].
ARAT consists of 4 domains: domain 1 and domain 3 tasks consists of grasping and
pinching various size of objects such as wooden blocks or marbles and then moving
them to top of the wooden box by reaching movement. Domain 2 mainly involves
moving items on a table focusing on grip function, and domain 4 items are gross
movement tasks that require lifting the arm to the head or face [58].
Validation of the IMU-based motion analysis system used in this study showed that
the accuracy and reliability of the sensors themselves are very high regarding angles.
However, in the form of upper body and extremity wearable multi-sensor system, it
is impossible to move a single joint alone, but all joints systemically move in 3-
dimension including body trunk and contralateral upper extremity. Intra-subject and
inter-subject covariance was calculated for forearm supination / pronation and elbow
flexion / extension to evaluate the system reliability regarding gyrosensor derived
77
angular values and the range was acceptable considering that the reaching tasks were
not identical in terms of posture and the target point of reaching was not exactly
determined. For the position data derived from the accelerometers, we compared
calculated data in y and z direction with the estimated real moving distance measured
by a ruler. The calculated distance data was similar to that of the measured data, and
variability was also acceptable. Also, the calculated workspace and ROM during
ARAT were similar between the two extremities with no significant difference (Table
2). This may also support the reliability of the system derived parameter values. While
it is difficult to say that it provides a completely accurate measure, but it seems
reasonable to consider this system as providing consistent and meaningful data.
The workspace of right and left hand was mostly similar, since the ARAT repeat the
same tasks with both hands alternatively. The slight difference between both sides
would be probably due to the difference in posture and orientation by limb dominance.
During the ADL tasks, the workspace of the dominant hand, which was right hand in
all subjects, was significantly larger than the non-dominant side by up to nearly 20cm
for all directions. In the view of stroke rehabilitation, most of the patients demonstrate
hemiplegia up to over 80% [59], which means that the intact limb should be able to
perform all normal functions. Patients with hemiplegia would use their intact hand as
their dominant hand, therefore, in some occasions the exoskeleton may only need to
cover smaller workspace than the dominant side.
The ROM of major upper extremity joints during essential daily activities are
presented in Table 2. Forearm supination / pronation and elbow flexion / extension
78
showed the highest values for the dominant side. The ROM for forearm supination /
pronation was 128.09° and 108.00° in average for right and left sides during all ADL
tasks. In a study performed with reflective marker based motion capture system, the
whole ROM calculated by overlapping all 95% confidential interval range during
various ADL tasks was 92° [26]. Another study done with electromagnetic sensor
system reported at the maximal supination angle from full pronation was 110° during
glass drinking and 75° while combing hair [25]. A study by van Andel et al. [60]
evaluated 4 selected ADL tasks with optic marker based system and their reported
ROM for forearm supination / pronation was approximately 130°. Regarding elbow
flexion / extension ROM, other studies also showed similar results. Aizawa et al. [25]
reported approximately 120° to 130° of ROM during various tasks, and Gates et al.
[26] showed that peak flexion angle of the elbow joint was 121° in average during
drinking from a cup, which was the highest value among the evaluated tasks. Another
study reported an ROM of around 140° from full extension [60]. Wrist dorsiflexion /
volarflexion ROM was also similar with other studies which ranged from 90° to 130°,
where in our study it was 113.70° and 110.08° for right and left side, respectively. It
would be important to ensure sufficient ROM for elbow flexion / extension, forearm
supination / pronation, and wrist dorsiflexion / volarflexion movements during
rehabilitation, because these joint movements are essential for performing ADL tasks
while recovery for distal joints are relatively slow and not sufficient for a large portion
of stroke patients [61-63].
The reason for evaluating major joint angular change during reaching after grasping
79
/ pinching was that these actions are most basic and at the same time most important
movements for performing any kind of tasks [64, 65], and most of the activities are
performed within the spatial range of these actions. Pinching was performed at a
slightly but significantly more abducted and flexed posture of the shoulder joint, and
also showed significant difference in fine tuning movements of the wrist joint. During
the reaching movement, the elbow joint was extended for nearly 70° from almost full
flexed position, and the forearm was rotated toward pronation direction for more than
36° from its initial supinated posture. In the current motion capture system, forearm
supination / pronation and wrist rotation are given separately, and it is reasonable to
assume that the sum of both forearm and wrist rotation would correspond to gross
supination / pronation angle. Therefore, it seems that the extent of forearm rotation
angle during reaching movement would reach near 45° in average.
In contrast to simple pure reaching movement, reaching movement associated with
performing a task may differ significantly regarding arm postures, grasping position,
and orientation [66, 67]. Human motor system has high redundancy in terms of multi-
degree-of-freedom control system, and while task-relevant factors are specifically
controlled, task-irrelevant variables are given relatively high variability [66]. In this
study, shoulder abduction / adduction and flexion / extension angles showed
significantly different posture between grasping and pinching, which reflects
different position of the elbow joint while performing the task. Wrist deviation and
rotation angles also showed significant difference reflecting difference in fine motor
posture and movements. Given the difference in posture, the main components of the
80
reaching movement: elbow flexion / extension and forearm supination / pronation did
not significantly differ between the two types of tasks. This result may be applied to
the swivel angle model suggested by Li et al. [66], where the shoulder joint angles
can be simplified to a swivel angle regarding the orientation and posture, and the
other distal joint angles account for essential reaching movements. In regular stroke
rehabilitation, proximal muscle power recovery occurs in the early stage and more
sufficiently compared to distal muscles [61-63], so it would be reasonable to motivate
the patient to practice taking an appropriate posture for providing the right orientation
of the exoskeleton (or upper extremity) using the proximal muscles voluntarily with
the help of gravity support system, while the individual robot joint actuation focus on
essential distal joint movements such as elbow flexion / extension, forearm supination
/ pronation and wrist movements.
For the workspace during ADL tasks and logsum / time (rate of displacement) during
many tasks, healthy subjects and stroke patients showed significant differences in
average, meaning that the workspace of hemiplegic patients is smaller, and the
movement speed is also slower. However, logsum (accumulated displacement) itself
had a rather larger value in stroke patients, suggesting that there may have been more
jerky movements in patients. In the current experiment, the stroke subjects were
instructed to complete all tasks with assistance if necessary, because it was
hypothesized that even with assistance there would be some extent of postural
difference regarding the angular range of the joints between healthy subjects and
stroke patients, and also between mildly and severely impaired patients.
81
Most of the parameters that showed significant difference between healthy subjects
and stroke patients and also significant correlation with clinical measures were
average amplitude of motion segments throughout the tasks. This implies that the
smoothness and voluntary movement magnitude get improved throughout the
recovery process of stroke, and it may serve as a useful clinical outcome measure if
simply accessible. If a wearable wrist sensor with an accelerometer and gyrosensor
can access some of these parameters such as forearm supination / pronation average
amplitude, it may provide clinically relevant data without the difficulty of wearing a
suit system or taking patients to a motion lab. There have been many investigations
regarding possible parameters from wearable sensors [68], but to our knowledge, the
average amplitude during movements has not been sufficiently investigated in the
clinical view.
82
4.4 Usability Test for an Image-processing Based Hand
Rehabilitation Robot
The usability test results showed high scores for interest, motivation, and safety issues,
but relatively low scores for difficulty, comfort, and expectance of improvement.
Because this system did not use a computer display as the main interface but instead
real objects, and that this may be used at home or at the bedside, the respondents
replied that these features were interesting and that the robot may help patients be
motivated for the therapy. The response from the patient group showed lower scores
in most of the categories except safety. The patients responded that the device was
not appropriate for their stroke recovery stage; however, they also commented that
the device would be very useful for the patients that are not able to move the distal
upper limb. The main issues from the free comments were that the gripping was not
highly secure, and that the task was limited due to the low degree of freedom. At
initial design, a height adjustment system was proposed so that the robot may grasp
objects at various heights; however, for the current prototype it was not applied due
to the structural complexity. In the next version, the height adjustment system using
a gravity compensation method is planned to be applied. Another feature of the
current prototype was that the hand part was placed at the palm side of the hand, in
contrast to other commercial hand exoskeletons where it is place at the dorsum of the
hand. The reason for this design was to prevent injury since it is difficult to control
gripping pressure. However, in this prototype, it was a problem in that it was not easy
for the user to determine if the object was sufficiently grasped so that it would not
83
fall down. The structure and position of the motors seemed to interfere with the user’s
workspace, which should be modified in the next step prototype. The use of pressure
sensors should be considered in the next step as well. In addition, a clinical proof-of-
concept study should be performed, which was not available during this study due to
IRB approval and FDA clearance issues. It is necessary that the investigational device
exemption (IDE) for medical robots with non-significant risk to be practically applied
in Korea, to facilitate the development and the proof-of-concept clinical studies for
medical robots, so that the clinically relevant robots may enter the market and be used
in clinics as soon as possible, whereas clinically irrelevant robots may stop
development at an earlier stage.
84
4.5 Optimization of Neurorehabilitation Robot Design Regarding
Clinical Settings
The purpose of this thesis was to evaluate and identify clinically relevant biomedical
factors, to eventually develop a practical rehabilitation exoskeleton robot. It was not
possible to evaluate clinical effects of the hand rehabilitation robot due to IRB and
FDA approval issues, therefore, it is necessary to discuss the estimated expected
therapeutic effects of the suggested robot based on previous clinical studies, and also
suggest a direction for overall design and development of the rehabilitation robots
with the viewpoint of optimization.
It is reported that minimal clinically important difference (MCID) for the Fugl-Meyer
(F-M) upper extremity motor score is estimated to range from 4.25 to 7.25 points [69],
whereas another study suggests 9 to 10 points [70]. In a multi-center randomized
controlled clinical trial using the ARMin exoskeleton, it was reported that the robotic
therapy group showed a 3.4 point improvement at the end of 8-week therapy
compared to the initial score, and it was approximately 0.78 points larger than the
conventional therapy group [52]. In the ARMin study, the stroke patients were mostly
in the chronic stage, and they had received the robotic therapy for 45 minutes per
session, three times a week, over 8 weeks. The weekly treatment dose was 2.25 hours,
and the total treatment dose was 18 hours of robotic therapy. From the individual
study data of the Cochrane review published in 2015, the total robotic treatment dose
and improvement in F-M upper extremity motor score showed linear correlation as
in Equation 4, with r2 = 0.57, as shown in Fig. 15 [2, 39, 40, 52, 71-73].
85
Fig. 15. Improvement in Fugl-Meyer upper extremity score after robot-assisted
treatment showed linear correlation with total treatment dose.
86
Fugl-Meyer Score Change (improvement)
= Total robot-assisted treatment dose (hours) × 0.073 + 1.680 (4)
In an intensive rehabilitation unit or clinic, one rehabilitation robot device can provide
approximately 16 thirty-minute sessions assuming that it is run for 8 hours daily.
However, for one patient, the robotic therapy should be provided at least 2 hours per
day to achieve the sufficient repetition level, which an occupational therapist cannot
provide to a single patient. A clinical study in Mexico showed that robot-assisted gait
training for 24 two-hour sessions resulted in significant improvement in lower
extremity F-M motor score compared to the control group consisting of 30-minute
sessions [74]. In this case, one robot can be used by only four patients per day. Two
hours per day would give 10 hours of weekly dose, and assuming linearity of
correlation for the total treatment dose and the F-M motor score improvement to
calculate ideally maximal expectation for improvement, it would be approximately
12 points increase of F-M upper extremity motor score after 3 months (12 weeks) of
robotic therapy.
However, most of the rehabilitation robots used in earlier clinical studies did not
include assistance for hand or finger movements, and therefore did not show
significant improvements in F-M scoring of the hand and fingers, while they account
for 14 points of the 66 points in the upper extremity (Supplement 4). Recently, there
have been several clinical trials using robotic devices focusing on hand and finger
recovery, with sufficient total treatment dose: 5 days a week for a duration of 3 to 8
weeks [75-78]. In these clinical studies, improvements in F-M score of distal limbs
87
ranged from 3 to 6 points [75-78]. It is shown in various clinical studies that the
improvement is generally specific to the joints targeted by the robotic therapy [79]. A
recent study using InMotionTM combined with task-specific training, for 90 to 100
minutes per day for 5 days a week over a 4-week period, showed an improvement of
7.7 points, where proximal F-M scores increased by 4.3 points and distal scores were
improved by 3.4 points [80]. Therefore, it may be assumed that an optimized robotic
therapy including hand and finger movements for a sufficient period of time would
expect nearly 18-point increase in F-M upper extremity motor score, which is
obviously a clinically significant improvement.
In contrast to optimization of the industrial robots, which are optimized for pre-
determined specific tasks only, rehabilitation robots require flexibility of tasks. As
evaluated in this study, it is necessary to determine workspace, ROM, and degree of
freedom as eventual specifications of the robot, rather than specific limited tasks,
although the workspace and ROM should be calculated from the desired tasks, among
the top ranked survey ADLs.
To apply the concept of optimization in rehabilitation robots, it should be assessed
with the viewpoint of maximizing task-specific repetition, especially to the range that
conventional therapy may not achieve. In addition, it should be considered that each
patient should be provided with maximal repetition as much as possible, but not just
purchasing one rehabilitation robot in the clinic and utilizing it for the whole working
day.
88
As mentioned previously, the purpose of developing a rehabilitation therapy robot is
not just substituting occupational therapy or adding a 30-minute treatment session
daily. Moreover, the algorithm and control method of the robot itself may not
influence the brain recovery better than conventional therapy as long as the therapy
consists of task-specific movements. Therefore, the optimal robot should be designed
in a self-usable form with or without help from the caregiver, but without the therapist,
as well as in a wearable and portable form. The weight issue should be discussed at
this point. Based on our design experience considering appropriate output torque, size,
and velocity, the weight of the motor and gear part together for each mechanical joint
ranges from approximately 0.5 kg in distal joints to 1.0 kg in middle joints such as
elbow flexion / extension, and to up to 2.0 kg for proximal joints such as shoulder
flexion / extension and abduction / adduction. The hand rehabilitation robot
developed in this study weighed 3.9 kg in total including the rolling bottom board
(1.2 kg), and if the metal plates for support were removed, the exoskeleton weight
itself was about 2.0 kg. For the patients or caregivers to freely move the robots from
one place to another and install it on the table, it seems that 4 to 5 kg is the maximum
tolerable weight. The KNRC self-feeding robot, which is a robot-arm type feeding
assistant robot, weighed approximately 3.7 to 4.7 kg for their pre-market prototype
[81]. These robots are not light enough to be lifted easily, but they are portable and
movable within home settings.
Even after adequate development of a rehabilitation robot, there still exist a number
of barriers for the robot to enter the market or the clinic: which may be classified into
89
technological, behavioral, organizational, and economic barriers [82]. The device
must be easily controllable for the patients and / or caregivers to let them adopt the
use of the robot successfully [82], and strong clinical evidence for the efficacy of the
neurorehabilitation robots should be established, since the evidence up to now is
relatively weak [2]. In addition, the healthcare system should be adequately modified
and also support the distribution and utilization of the robot for successful initial
adoption, especially when the cost-effectiveness is not well established [82].
There was a study that calculated healthcare cost for comparison of a robot therapy
group, an intensive therapy group, and a usual therapy group in the VA-ROBOTICS
study using the MIT-Manus robot [72, 83]. The study showed that the intervention
cost for a single session was lower for robot therapy ($140) compared to the intensive
therapy ($218), and the total healthcare cost at 36 weeks showed approximately
$2,000 in savings for the robotic therapy group compared to intensive therapy.
However, the F-M score showed only 2.17 points difference, which may not have a
clinically significant meaning.
Assuming full utilization of the robot in the hospital, 17 patients are able to use the
Armeo® type robot during a week, and 4 patients can use the practical robot during a
week, which is 2 hours daily per patient [74]. Assuming a rehabilitation clinic can
afford to purchase one Armeo® Power ($190,000), which approximately 17 patients
may use weekly, the practical robot price should be kept under $50,000 per robot to
provide robotic therapy for a similar number of patients.
90
Summarizing this discussion section, the clinically relevant practical rehabilitation
exoskeleton for treatment in a hospital should have at most 3 axes (or 2 axes + hand
part) distal to the elbow, with proximal structures supported by gravity compensation,
with total movable weight under 5 kg, and a final market price under $50,000, which
is operable by the patient and / or the caregiver, to provide maximal task-specific
repetition in the inpatient rehabilitation setting.
4.6 Limitations
This study has several limitations. For the demand survey, the number of subjects was
not sufficient to generalize the needs of all upper extremity impaired persons. UIG
was relatively homogeneous and representative of chronic stroke patients, however,
BIG was mostly consisted of young muscular dystrophy patients and only 5 cervical
spinal cord injured patients, potentially possessing limitations for generalization.
However, most previous studies were performed with spinal cord injury or ALS
patients. Therefore, this study may provide a reference for comparison between
different disease entities. And most of the participants were not familiar with the
concept of a BMI other than the explanation given to them just before the survey.
There may have been some difficulties in imaging how it would be with the given
technology and functions. If they had known more about BMIs and rehabilitation
robots, the survey results would have been more accurate.
For the spasticity resistance study, the clinical assessment of spasticity with the MAS
91
grading system was not clearly applicable for some subjects. As most of the
volunteers were chronic stroke patients with an average of 9.7 years since stroke onset,
the characteristics of their spasticity were very complex and did not typically fit to
the definition of the MAS grade. For this study, two experienced physiatrists
evaluated spasticity independently and there were disagreements between them
concerning four subjects for the elbow and five subjects for the wrist. In those cases,
the higher grade was taken for analysis. The trials were performed from a slow speed
to a fast speed, in order to minimize the risk of musculoskeletal injury of the joint.
However, preceding trials at the slower speeds could have decreased the spasticity
temporarily, and the resistance torques for the faster speeds may have been
underestimated. Ideally, randomization of the trial order in terms of angular velocity
would have resulted in more accurate data. In addition, the numbers of subjects at
each level of spasticity were not sufficient to generalize the results of the study
For the motion analysis, the number of subjects was relatively small to generalize the
findings of motion analysis. However, the statistical analyses provided minimal
requirements regarding validity and reliability. IMU-based sensors basically have its
inevitable limitations, which include drift phenomenon in both position and angular
values, and it would have affected the outcome measure values [84]. Also gimbal-
lock phenomenon regarding especially shoulder joint angles may have occurred
during data measurements [25]. In this point of view, the data may not be accurate in
terms of absolute values, however, it seems that the general pattern of the data is
reliable since the data are sufficiently consistent.
92
The hand rehabilitation robot prototype also had a number of limitations. The height
adjustment system was not applied resulting in limited function of the robot. The
contour of the hand part need to be more customized to the real contour of the user’s
hand to provide a better sense of grabbing objects. This may be solved by using 3-
dimensional printers.
93
5. CONCLUSIONS AND FUTURE WORK
The results of this research will serve as a basis for the design and development of a
practical and portable but clinically relevant neurorehabilitation exoskeleton robot.
Clinical evidence should be supported for pilot developments to successfully enter
the clinical market and become widely used among stroke patients.
The next step after this study would be developing a simple and portable
neurorehabilitation robot by applying the biomedical factors described in this study.
One suggestion could be estimating the elbow trajectory for operating in the essential
workspace and making an elbow support with a constant elastic spring for gravity
compensation and at the same time movable in 3-dimension within the intended
trajectory area. The part distal to the elbow part may be designed by accommodating
the described hand rehabilitation robot and minimizing the ROM according to this
study for a smaller and lightweight design.
For optimal control of the robot, biomechanical analysis regarding joint torque during
the essential movements may be necessary to provide support as needed and to avoid
over-actuation of a joint. Millard et al. [85] presented an example of optimal control
in lifting motion. Upper limb movement simulation using muscle-based modeling
software may provide useful information [86].
Further research on user-intent driven actuation of the exoskeleton must be performed
for better stimulation of neuroplasticity, using both brain signals and sensing of the
94
volitional movements in the peripheral limb. A clinical proof-of-concept study should
be performed prior to further developments to ensure the efficacy of the proposed
concept.
95
Acknowledgments
A part of this thesis has been published in the IEEE Transactions on Neural Systems
and Rehabilitation Engineering:
Nam HS, Koh S, Kim YJ, Beom J, Lee WH, Lee SU, Kim S. Biomechanical
Reactions of Exoskeleton Neurorehabilitation Robots in Spastic Elbows and Wrists.
IEEE Trans Neural Syst Rehabil Eng. 2017;25(11):2196-2203.
©2017 IEEE. In reference to IEEE copyrighted material which is used with
permission in this thesis, the IEEE does not endorse any of Seoul National
University’s products or services. Internal or personal use of this material is permitted.
However, permission to reprint or republish IEEE copyrighted material for
advertising or promotional purposes or for creating new collective works for resale
or redistribution, or to reuse any copyrighted component of this work in other works
must be obtained from the IEEE (See Appendix for IEEE permission statement).
96
Funding
This study was supported by a grant (NRCTR-EX15002, NRCTR-EX16008)
from the Translational Research Center for Rehabilitation Robots, Korea
National Rehabilitation Center, Ministry of Health & Welfare, Korea, the
Brain Fusion Program of Seoul National University (800-20120444), the
Brain Research Program through the National Research Foundation of
Korea(NRF) funded by the Ministry of Science, ICT & Future Planning
(2016M3C7A1904984), and the General Research Program of Seoul National
University Hospital (04-2016-0870).
97
REFERENCES
1. Carolei A, Sacco S, De Santis F, Marini C. Epidemiology of stroke. Clin Exp
Hypertens. 2002;24(7-8):479-83.
2. Mehrholz J, Pohl M, Platz T, Kugler J, Elsner B. Electromechanical and robot-
assisted arm training for improving activities of daily living, arm function, and arm
muscle strength after stroke. Cochrane Database Syst Rev. 2015(11):CD006876.
3. Huang VS, Krakauer JW. Robotic neurorehabilitation: a computational motor
learning perspective. J Neuroeng Rehabil. 2009;6:5.
4. Bertani R, Melegari C, De Cola MC, Bramanti A, Bramanti P, Calabro RS. Effects
of robot-assisted upper limb rehabilitation in stroke patients: a systematic review with
meta-analysis. Neurol Sci. 2017;38(9):1561-9.
5. Lo HS, Xie SQ. Exoskeleton robots for upper-limb rehabilitation: state of the art
and future prospects. Med Eng Phys. 2012;34(3):261-8.
6. Beom J, Koh S, Nam HS, Kim W, Kim Y, Seo HG, et al. Robotic Mirror Therapy
System for Functional Recovery of Hemiplegic Arms. J Vis Exp. 2016(114).
7. Nam HS, Koh S, Beom J, Kim YJ, Park JW, Koh ES, et al. Recovery of
Proprioception in the Upper Extremity by Robotic Mirror Therapy: a Clinical Pilot
Study for Proof of Concept. J Korean Med Sci. 2017;32(10):1568-75.
8. Kim YJ, Park SW, Yeom HG, Bang MS, Kim JS, Chung CK, et al. A study on a
robot arm driven by three-dimensional trajectories predicted from non-invasive
neural signals. Biomed Eng Online. 2015;14:81.
9. Bedaf S, Gelderblom GJ, De Witte L. Overview and Categorization of Robots
98
Supporting Independent Living of Elderly People: What Activities Do They Support
and How Far Have They Developed. Assist Technol. 2015;27(2):88-100.
10. Jarrasse N, Proietti T, Crocher V, Robertson J, Sahbani A, Morel G, et al. Robotic
exoskeletons: a perspective for the rehabilitation of arm coordination in stroke
patients. Front Hum Neurosci. 2014;8:947.
11. Collinger JL, Wodlinger B, Downey JE, Wang W, Tyler-Kabara EC, Weber DJ, et
al. High-performance neuroprosthetic control by an individual with tetraplegia.
Lancet. 2013;381(9866):557-64.
12. Kwak NS, Muller KR, Lee SW. A lower limb exoskeleton control system based
on steady state visual evoked potentials. J Neural Eng. 2015;12(5):056009.
13. Trompetto C, Marinelli L, Mori L, Pelosin E, Curra A, Molfetta L, et al.
Pathophysiology of spasticity: implications for neurorehabilitation. Biomed Res Int.
2014;2014:354906.
14. Nam HS, Park YG, Paik NJ, Oh BM, Chun MH, Yang HE, et al. Efficacy and
safety of NABOTA in post-stroke upper limb spasticity: a phase 3 multicenter,
double-blinded, randomized controlled trial. J Neurol Sci. 2015;357(1-2):192-7.
15. Sommerfeld DK, Gripenstedt U, Welmer AK. Spasticity after stroke: an overview
of prevalence, test instruments, and treatments. Am J Phys Med Rehabil.
2012;91(9):814-20.
16. Adams MM, Hicks AL. Spasticity after spinal cord injury. Spinal Cord.
2005;43(10):577-86.
17. Opheim A, Danielsson A, Alt Murphy M, Persson HC, Sunnerhagen KS. Upper-
limb spasticity during the first year after stroke: stroke arm longitudinal study at the
99
University of Gothenburg. Am J Phys Med Rehabil. 2014;93(10):884-96.
18. Bohannon RW, Andrews AW. Interrater reliability of hand-held dynamometry.
Phys Ther. 1987;67(6):931-3.
19. Sanger TD, Delgado MR, Gaebler-Spira D, Hallett M, Mink JW, Task Force on
Childhood Motor D. Classification and definition of disorders causing hypertonia in
childhood. Pediatrics. 2003;111(1):e89-97.
20. Nef T, Mihelj M, Riener R. ARMin: a robot for patient-cooperative arm therapy.
Med Biol Eng Comput. 2007;45(9):887-900.
21. Kumar RT, Pandyan AD, Sharma AK. Biomechanical measurement of post-stroke
spasticity. Age Ageing. 2006;35(4):371-5.
22. Keller U, Scholch S, Albisser U, Rudhe C, Curt A, Riener R, et al. Robot-assisted
arm assessments in spinal cord injured patients: a consideration of concept study.
PLoS One. 2015;10(5):e0126948.
23. Alibiglou L, Rymer WZ, Harvey RL, Mirbagheri MM. The relation between
Ashworth scores and neuromechanical measurements of spasticity following stroke.
J Neuroeng Rehabil. 2008;5:18.
24. Valero-Cuevas FJ, Santello M. On neuromechanical approaches for the study of
biological and robotic grasp and manipulation. J Neuroeng Rehabil. 2017;14(1):101.
25. Aizawa J, Masuda T, Koyama T, Nakamaru K, Isozaki K, Okawa A, et al. Three-
dimensional motion of the upper extremity joints during various activities of daily
living. J Biomech. 2010;43(15):2915-22.
26. Gates DH, Walters LS, Cowley J, Wilken JM, Resnik L. Range of Motion
Requirements for Upper-Limb Activities of Daily Living. Am J Occup Ther.
100
2016;70(1):7001350010.
27. Kim K, Song WK, Lee J, Lee HY, Park DS, Ko BW, et al. Kinematic analysis of
upper extremity movement during drinking in hemiplegic subjects. Clin Biomech
(Bristol, Avon). 2014;29(3):248-56.
28. Chen J, Lum PS. Pilot testing of the spring operated wearable enhancer for arm
rehabilitation (SpringWear). J Neuroeng Rehabil. 2018;15(1):13.
29. Downey JE, Weiss JM, Muelling K, Venkatraman A, Valois JS, Hebert M, et al.
Blending of brain-machine interface and vision-guided autonomous robotics
improves neuroprosthetic arm performance during grasping. J Neuroeng Rehabil.
2016;13:28.
30. Bang MS, Kim S, Nam HS, Kim YJ, Oh BM, Beom J, et al. Upper Limb
Rehabilitation Robot Module and Upper Limb Rehabilitation System for Precision
Control by Image Processing. Korea Patent 10-1831602. 2018.
31. Shah S, Vanclay F, Cooper B. Improving the sensitivity of the Barthel Index for
stroke rehabilitation. J Clin Epidemiol. 1989;42(8):703-9.
32. Chumney D, Nollinger K, Shesko K, Skop K, Spencer M, Newton RA. Ability of
Functional Independence Measure to accurately predict functional outcome of stroke-
specific population: systematic review. J Rehabil Res Dev. 2010;47(1):17-29.
33. Peng Q, Park HS, Shah P, Wilson N, Ren Y, Wu YN, et al. Quantitative evaluations
of ankle spasticity and stiffness in neurological disorders using manual spasticity
evaluator. J Rehabil Res Dev. 2011;48(4):473-81.
34. Collinger JL, Boninger ML, Bruns TM, Curley K, Wang W, Weber DJ. Functional
priorities, assistive technology, and brain-computer interfaces after spinal cord injury.
101
J Rehabil Res Dev. 2013;50(2):145-60.
35. Huggins JE, Wren PA, Gruis KL. What would brain-computer interface users
want? Opinions and priorities of potential users with amyotrophic lateral sclerosis.
Amyotroph Lateral Scler. 2011;12(5):318-24.
36. Huggins JE, Moinuddin AA, Chiodo AE, Wren PA. What would brain-computer
interface users want: opinions and priorities of potential users with spinal cord injury.
Arch Phys Med Rehabil. 2015;96(3 Suppl):S38-45 e1-5.
37. Watkins CL, Leathley MJ, Gregson JM, Moore AP, Smith TL, Sharma AK.
Prevalence of spasticity post stroke. Clin Rehabil. 2002;16(5):515-22.
38. Zorowitz RD, Gillard PJ, Brainin M. Poststroke spasticity: sequelae and burden
on stroke survivors and caregivers. Neurology. 2013;80(3 Suppl 2):S45-52.
39. Daly JJ, Hogan N, Perepezko EM, Krebs HI, Rogers JM, Goyal KS, et al.
Response to upper-limb robotics and functional neuromuscular stimulation following
stroke. J Rehabil Res Dev. 2005;42(6):723-36.
40. Conroy SS, Whitall J, Dipietro L, Jones-Lush LM, Zhan M, Finley MA, et al.
Effect of gravity on robot-assisted motor training after chronic stroke: a randomized
trial. Arch Phys Med Rehabil. 2011;92(11):1754-61.
41. McCabe J, Monkiewicz M, Holcomb J, Pundik S, Daly JJ. Comparison of robotics,
functional electrical stimulation, and motor learning methods for treatment of
persistent upper extremity dysfunction after stroke: a randomized controlled trial.
Arch Phys Med Rehabil. 2015;96(6):981-90.
42. Anderson KD. Targeting recovery: priorities of the spinal cord-injured population.
J Neurotrauma. 2004;21(10):1371-83.
102
43. de Leva P. Adjustments to Zatsiorsky-Seluyanov's segment inertia parameters. J
Biomech. 1996;29(9):1223-30.
44. Park HS, Kim J, Damiano DL. Development of a Haptic Elbow Spasticity
Simulator (HESS) for improving accuracy and reliability of clinical assessment of
spasticity. IEEE Trans Neural Syst Rehabil Eng. 2012;20(3):361-70.
45. Ren Y, Kang SH, Park HS, Wu YN, Zhang LQ. Developing a multi-joint upper
limb exoskeleton robot for diagnosis, therapy, and outcome evaluation in
neurorehabilitation. IEEE Trans Neural Syst Rehabil Eng. 2013;21(3):490-9.
46. Starsky AJ, Sangani SG, McGuire JR, Logan B, Schmit BD. Reliability of
biomechanical spasticity measurements at the elbow of people poststroke. Arch Phys
Med Rehabil. 2005;86(8):1648-54.
47. Kung PC, Lin CC, Ju MS. Neuro-rehabilitation robot-assisted assessments of
synergy patterns of forearm, elbow and shoulder joints in chronic stroke patients. Clin
Biomech (Bristol, Avon). 2010;25(7):647-54.
48. Malhotra S, Cousins E, Ward A, Day C, Jones P, Roffe C, et al. An investigation
into the agreement between clinical, biomechanical and neurophysiological measures
of spasticity. Clin Rehabil. 2008;22(12):1105-15.
49. Herrnstadt G, Alavi N, Randhawa BK, Boyd LA, Menon C. Bimanual elbow
robotic orthoses: preliminary investigations on an impairment force-feedback
rehabilitation method. Front Hum Neurosci. 2015;9:169.
50. Looned R, Webb J, Xiao ZG, Menon C. Assisting drinking with an affordable
BCI-controlled wearable robot and electrical stimulation: a preliminary investigation.
J Neuroeng Rehabil. 2014;11:51.
103
51. Pirondini E, Coscia M, Marcheschi S, Roas G, Salsedo F, Frisoli A, et al.
Evaluation of the effects of the Arm Light Exoskeleton on movement execution and
muscle activities: a pilot study on healthy subjects. J Neuroeng Rehabil. 2016;13:9.
52. Klamroth-Marganska V, Blanco J, Campen K, Curt A, Dietz V, Ettlin T, et al.
Three-dimensional, task-specific robot therapy of the arm after stroke: a multicentre,
parallel-group randomised trial. Lancet Neurol. 2014;13(2):159-66.
53. Seth N, Johnson D, Taylor GW, Allen OB, Abdullah HA. Robotic pilot study for
analysing spasticity: clinical data versus healthy controls. J Neuroeng Rehabil.
2015;12:109.
54. Wu YN, Ren Y, Goldsmith A, Gaebler D, Liu SQ, Zhang LQ. Characterization of
spasticity in cerebral palsy: dependence of catch angle on velocity. Dev Med Child
Neurol. 2010;52(6):563-9.
55. Hsieh CL, Hsueh IP, Chiang FM, Lin PH. Inter-rater reliability and validity of the
action research arm test in stroke patients. Age Ageing. 1998;27(2):107-13.
56. Koh CL, Hsueh IP, Wang WC, Sheu CF, Yu TY, Wang CH, et al. Validation of the
action research arm test using item response theory in patients after stroke. J Rehabil
Med. 2006;38(6):375-80.
57. Chen HF, Lin KC, Wu CY, Chen CL. Rasch validation and predictive validity of
the action research arm test in patients receiving stroke rehabilitation. Arch Phys Med
Rehabil. 2012;93(6):1039-45.
58. Yozbatiran N, Der-Yeghiaian L, Cramer SC. A standardized approach to
performing the action research arm test. Neurorehabil Neural Repair. 2008;22(1):78-
90.
104
59. Parker VM, Wade DT, Langton Hewer R. Loss of arm function after stroke:
measurement, frequency, and recovery. Int Rehabil Med. 1986;8(2):69-73.
60. van Andel CJ, Wolterbeek N, Doorenbosch CA, Veeger DH, Harlaar J. Complete
3D kinematics of upper extremity functional tasks. Gait Posture. 2008;27(1):120-7.
61. Krakauer JW. Arm function after stroke: from physiology to recovery. Semin
Neurol. 2005;25(4):384-95.
62. Shelton FN, Reding MJ. Effect of lesion location on upper limb motor recovery
after stroke. Stroke. 2001;32(1):107-12.
63. Brunnstrom S. Motor testing procedures in hemiplegia: based on sequential
recovery stages. Phys Ther. 1966;46(4):357-75.
64. Shumway-Cook A, Woollacott MH. Motor Control: Translating Research into
Clinical Practice. Fifth Edition. Philadelphia, PA: Wolters Kluwer; 2017.
65. Guccione AA, Wong RA, Avers D. Geriatric Physical Therapy. 3rd Edition. St.
Louis, Missouri: Elsevier; 2012.
66. Li Z, Milutinovic D, Rosen J. From reaching to reach-to-grasp: the arm posture
difference and its implications on human motion control strategy. Exp Brain Res.
2017;235(5):1627-42.
67. Scharoun SM, Scanlan KA, Bryden PJ. Hand and Grasp Selection in a Preferential
Reaching Task: The Effects of Object Location, Orientation, and Task Intention. Front
Psychol. 2016;7:360.
68. Tran VD, Dario P, Mazzoleni S. Kinematic measures for upper limb robot-assisted
therapy following stroke and correlations with clinical outcome measures: A review.
Med Eng Phys. 2018;53:13-31.
105
69. Page SJ, Fulk GD, Boyne P. Clinically important differences for the upper-
extremity Fugl-Meyer Scale in people with minimal to moderate impairment due to
chronic stroke. Phys Ther. 2012;92(6):791-8.
70. Arya KN, Verma R, Garg RK. Estimating the minimal clinically important
difference of an upper extremity recovery measure in subacute stroke patients. Top
Stroke Rehabil. 2011;18 Suppl 1:599-610.
71. Brokaw EB, Nichols D, Holley RJ, Lum PS. Robotic therapy provides a stimulus
for upper limb motor recovery after stroke that is complementary to and distinct from
conventional therapy. Neurorehabil Neural Repair. 2014;28(4):367-76.
72. Lo AC, Guarino PD, Richards LG, Haselkorn JK, Wittenberg GF, Federman DG,
et al. Robot-assisted therapy for long-term upper-limb impairment after stroke. N
Engl J Med. 2010;362(19):1772-83.
73. Wu X, Guarino P, Lo AC, Peduzzi P, Wininger M. Long-term Effectiveness of
Intensive Therapy in Chronic Stroke. Neurorehabil Neural Repair. 2016;30(6):583-
90.
74. Bustamante Valles K, Montes S, Madrigal Mde J, Burciaga A, Martinez ME,
Johnson MJ. Technology-assisted stroke rehabilitation in Mexico: a pilot randomized
trial comparing traditional therapy to circuit training in a Robot/technology-assisted
therapy gym. J Neuroeng Rehabil. 2016;13(1):83.
75. Rowe JB, Chan V, Ingemanson ML, Cramer SC, Wolbrecht ET, Reinkensmeyer
DJ. Robotic Assistance for Training Finger Movement Using a Hebbian Model: A
Randomized Controlled Trial. Neurorehabil Neural Repair. 2017;31(8):769-80.
76. Vanoglio F, Bernocchi P, Mule C, Garofali F, Mora C, Taveggia G, et al.
106
Feasibility and efficacy of a robotic device for hand rehabilitation in hemiplegic
stroke patients: a randomized pilot controlled study. Clin Rehabil. 2017;31(3):351-
60.
77. Orihuela-Espina F, Roldan GF, Sanchez-Villavicencio I, Palafox L, Leder R,
Sucar LE, et al. Robot training for hand motor recovery in subacute stroke patients:
A randomized controlled trial. J Hand Ther. 2016;29(1):51-7; quiz 7.
78. Susanto EA, Tong RK, Ockenfeld C, Ho NS. Efficacy of robot-assisted fingers
training in chronic stroke survivors: a pilot randomized-controlled trial. J Neuroeng
Rehabil. 2015;12:42.
79. Veerbeek JM, Langbroek-Amersfoort AC, van Wegen EE, Meskers CG, Kwakkel
G. Effects of Robot-Assisted Therapy for the Upper Limb After Stroke. Neurorehabil
Neural Repair. 2017;31(2):107-21.
80. Hung CS, Hsieh YW, Wu CY, Lin YT, Lin KC, Chen CL. The Effects of
Combination of Robot-Assisted Therapy With Task-Specific or Impairment-Oriented
Training on Motor Function and Quality of Life in Chronic Stroke. PM R.
2016;8(8):721-9.
81. Song WK, Song WJ, Kim Y, Kim J. Usability test of KNRC self-feeding robot.
IEEE Int Conf Rehabil Robot. 2013;2013:6650501.
82. Turchetti G, Vitiello N, Trieste L, Romiti S, Geisler E, Micera S. Why
effectiveness of robot-mediated neurorehabilitation does not necessarily influence its
adoption. IEEE Rev Biomed Eng. 2014;7:143-53.
83. Wagner TH, Lo AC, Peduzzi P, Bravata DM, Huang GD, Krebs HI, et al. An
economic analysis of robot-assisted therapy for long-term upper-limb impairment
107
after stroke. Stroke. 2011;42(9):2630-2.
84. Luinge HJ, Veltink PH. Measuring orientation of human body segments using
miniature gyroscopes and accelerometers. Med Biol Eng Comput. 2005;43(2):273-
82.
85. Millard M, Sreenivasa M, Mombaur K. Predicting the Motions and Forces of
Wearable Robotic systems Using optimal Control. Front Robot Ai. 2017;4.
86. Delp SL, Anderson FC, Arnold AS, Loan P, Habib A, John CT, et al. OpenSim:
open-source software to create and analyze dynamic Simulations of movement. IEEE
T Bio-Med Eng. 2007;54(11):1940-50.
108
Supplemental Materials
Supplement 1. Survey for potential users’ demand on assistive robots
* This survey is to be performed by the examiner throughout the whole process to explain each questions and activities of daily living items, and ensure that the respondent is always aware of the rating scale.
0. Please check your gender and age: M / F / Age: _____
1. Please check on your disease category that caused your impairment.
1) Stroke
2) Spinal cord injury
3) Muscular dystrophy
4) Motor neuron disease
5) Peripheral nerve injury
6) Any others: ____________________________
2. When was the onset of the impairment? Year ___________ Month ____ ( ____ years ago)
3. How well do you use your upper extremities?
1) I can use both arms functionally (at least partially)
2) I only use one arm functionally
3) I can hardly use both arms functionally
4. Please check on your gait status.
1) I can walk independently without any assistive tools.
2) I can walk independently using some assistive tools.
3) I need other person’s assistance (regardless of assistive tool use)
4) I hardly can walk despite any kind of help from others.
5. Is wheelchair your main method of moving? Y / N
* As you have been informed during your consent, please assume that the following robots (external robotic arm and upper limb exoskeleton: shown in pictures) may be controlled perfectly according to your intent, to perform following activities of daily living tasks. Regarding each task, please reply of your 1) current dependence on others, your ratings on 2) objective importance of the function from the viewpoint of a developer based on your experience with severe functional impairment, and 3) subjective necessity of the function from the viewpoint of a consumer based on your current daily activities. For the subjective necessity, please think if you would use the function if a perfect robot was provided for the specific function. Please rate objective importance and subjective necessity for the external robotic arm and upper limb exoskeleton separately.
109
*Ratings
5-Likert scale 1 2 3 4 5
Dependence Totally dependent
Mostly dependent
Half dependent
Mostly dependent
Totally dependent
Importance Unimportant Of little importance
Moderately important Important Very
important
Necessity Not necessary Of little necessity
Moderately necessary Necessary Highly
necessary
ADL items External Robotic Arm
Upper Limb Exoskeleton
Washing face Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Brushing teeth (including squeezing toothpaste)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Hairdressing Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Dressing (putting shirts on and off)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Eating Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Handling foods (i.e. peeling a banana, opening a bottle cap, etc.)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Cleaning (cleaning one’s desk) Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Moving close items Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Smartphone (using a smartphone or a tablet)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Computer (using a computer: keyboard and mouse)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Phone calls (dialing and receiving a phone call)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Writing Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
110
Switch control Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Purse (putting in and taking out bills and cards from a purse/wallet)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 1 2 3 4 5 Necessity 1 2 3 4 5 1 2 3 4 5
Transfer (assisting bed to chair, chair to standing, etc.)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 Necessity 1 2 3 4 5
Toilet use Dependency 1 2 3 4 5 Importance 1 2 3 4 5 Necessity 1 2 3 4 5
Self-exercise (of the upper extremity)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 Necessity 1 2 3 4 5
Wheelchair control (both manual and electric)
Dependency 1 2 3 4 5 Importance 1 2 3 4 5 Necessity 1 2 3 4 5
111
Supplement 2. Brunnstrom Stage [64]
1) Brunnstrom stage for arm
Stage Arm 1 Flaccidity-no voluntary movement 2 Synergies developing-flexion usually develops before extension
(may be a weak associated reaction or voluntary contraction with or without joint motion); spasticity developing
3 Synergies performed voluntarily Increased spasticity which may become marked
4 Some movements deviating from synergy a. Hand behind body b. Arm to forward-horizontal position c. Pronation-supination with elbow flexed to 90。; spasticity decreasing
5 Independence from the basic synergies a. Arm to side-horizontal position b. Arm forward and overhead c. Pronation-supination with elbow full extended; spasticity waning
6 Isolated joint movements freely performed with near normal coordination Spasticity minimal
112
2) Brunnstrom stage for hand
Stage Hand 1 Flaccidity 2 Little or no active finger flexion 3 Mass grasp or hook grasp
No voluntary finger extension or release 4 Lateral prehension with release by thumb movement
Semivoluntary finger extension (small range of motion) 5 Palmar prehension
Possible cylindrical and spherical grasp (awkward) Voluntary mass finger extension (variable range of motion)
6 All types of prehension (improved skill) Voluntary finger extension (full range of motion) Individual finger movements
113
Supplement 3. Action Research Arm Test Tasks [56]
Task Score* Time (sec) Left Right Left/Right A. Grasp (grasp and reach out to top of the shelf) 1. Block, 10 cm3 0 1 2 3 0 1 2 3 2. Block, 2.5 cm3 0 1 2 3 0 1 2 3 3. Block, 5 cm3 0 1 2 3 0 1 2 3 4. Block, 7.5 cm3 0 1 2 3 0 1 2 3 5. Cricket ball 0 1 2 3 0 1 2 3 6. Sharpening stone 0 1 2 3 0 1 2 3
Subtest score /18 /18 B. Grip 7. Pour water from glass to glass 0 1 2 3 0 1 2 3 8. Tube 2.25 cm 0 1 2 3 0 1 2 3 9. Tube 1 cm 0 1 2 3 0 1 2 3 10. Put washer over a bolt 0 1 2 3 0 1 2 3
Subtest score /12 /12 C. Pinch (pinch and reach out to top of the shelf) 11. Ball 6 mm 3rd finger and thumb 0 1 2 3 0 1 2 3 12. Marble 1st finger and thumb 0 1 2 3 0 1 2 3 13. Ball 6 mm 2nd finger and thumb 0 1 2 3 0 1 2 3 14. Ball 6 mm 1st finger and thumb 0 1 2 3 0 1 2 3 15. Marble 3rd finger and thumb 0 1 2 3 0 1 2 3 16. Marble 2nd finger and thumb 0 1 2 3 0 1 2 3
Subtest score /18 /18 D. Gross Movements 17. Hand behind head 0 1 2 3 0 1 2 3 18. Hand on top of head 0 1 2 3 0 1 2 3 19. Hand to mouth 0 1 2 3 0 1 2 3
Subtest score /9 /9 Total score /57 /57
* Scoring: 0 = unable to complete any part of the task within 60 sec; 1 = task partially performed within 60 sec; 2 = task completed but with great difficulty or abnormally long time; 3 = task completed normally within 5 sec
114
Supplement 4. Fugl-Meyer Assessment Scale for Upper Extremity Motor Function [59]
Upper Extremity Score (2/(1)/0) A. Shoulder/Elbow/Forearm I. Reflex activity Flexors Biceps, Finger flexors Extensors Triceps II. a. Flexor synergy Shouler Retraction Elevation Abduction Outward rotation Elbow Flexion Forearm Supination b. Extensor synergy Shoulder Adduction/inward rotation Elbow Extension Forearm Pronation III. Hand to lumbar spine Hand Move to lumbar spine Shoulder Flexion 0˚- 90˚ Elbow 90˚ Pronation/supination IV. Shoulder Abduction 0˚- 90˚ Flexion 90˚- 180˚ Elbow 0˚ Pronation/supination V. Normal reflex activity Total – Shoulder/Elbow/Forearm /36 B. Wrist Elbow 90˚ Wrist stability Elbow 90˚ Wrist flexion/extension Elbow 0˚ Wrist stability Elbow 0˚ Wrist flexion/extension Circumduction Total - Wrist /10 C. Hand Fingers mass flexion Fingers mass extension Grasp a Grasp b Grasp c Grasp d Grasp e Total – Hand /14 D. Coordination/Speed Tremor Dysmetria Speed Total – Coordination/Speed /6 Total Motor Score for the Upper Extremity /66
115
Appendix
116
국문초록
실용적인 상지 외골격 재활 로봇 설계 및 개발을
위한 임상 적합성 기반 의공학적 인자
서울대학교 의과대학 의학과
의공학교실 남형석
서론: 지난 수년간 재활 로봇은 빠른 속도로 발전되어 다양한 기능을
갖추고 실제 임상에 적용되고 있는 사례도 많다. 그러나 로봇의 본래
목적인 과제 지향적 반복적 움직임을 최대한 제공하여 기존의 치료
방식에 비해 유의하게 기능 회복과 뇌가소성을 촉진시킨 사례는 거의
없다고 알려져 있다. 본 연구의 목적은 단순한 인체공학적인 접근을 넘어,
실용적이면서도 간단한 외골격 뇌신경재활 로봇을 개발하기 위한
의공학적 인자를 정립하는 데에 있다.
방법: 우선 뇌졸중 및 신경근육질환으로 인한 상지 마비 환자 48 명을
대상으로 기술 수요 조사를 시행하였다. 로봇을 실제 사용하게 될 예비
사용자 대상의 설문 조사를 통해 로봇을 개발하는 실질적인 목적을
117
규명하고자 하였다. 중추신경계 손상 및 질환으로 인한 마비 환자는
특징적으로 경직 증상을 보이는 경우가 많다. 이에 경직 증상을 보이는
만성 뇌졸중 환자 20 명을 대상으로 하여, 경직이 있을 경우 외골격
로봇이 받게 되는 저항을 정량적으로 측정하였다. 이를 통해 외골격 로봇
주요 관절의 모터에 필요한 최소한의 토크 출력을 제시하고자 하였다.
건강한 자원자 10 명을 대상으로는 관성 측정 장치 기반 동작 분석
시스템을 이용하여 액션 리서치 암 테스트 및 상위 수요 일상 생활 동작
수행 시 손의 작업 공간 및 주요 관절의 가동 범위를 측정하였다. 같은
방법으로 브룬스트롬 3 단계에서 6 단계에 걸쳐 분포하는 뇌졸중 환자
9 명을 모집하여 편마비측 상지에서 보이는 동작 특성에 대해 분석하였다.
사용자 의도에 따른 로봇 제어를 위해 영상 처리 기반 제어 알고리즘을
제안하였고, 이를 이용한 시제품을 제작하여 의사, 공학자, 치료사 및
뇌졸중 환자를 대상으로 사용성 평가를 시행하였다.
결과: 예비 사용자 수요 조사 결과, 음식 다루기, 옷입기, 가까운 물건 옮
기기 등이 외골격 및 외부 로봇팔 모두에 가장 필요한 동작으로 나타났
다. 뇌졸중 환자의 경우 특히 로봇을 이용하여 자가 운동을 할 수 있기를
희망하였다. 팔꿈치 굴곡 및 손목 굴곡 경직으로 인한 로봇 저항은 경직
이 낮은 그룹 (수정 애쉬워스 척도 0, 1)에서 각각 3.68 ± 2.42, 4.23
± 1.75 Nm 이었고, 중간 경직 그룹 (1+)에서는 5.94 ± 2.55, 5.68
± 1.96 Nm, 높은 경직 그룹 (2, 3)에서는 8.25 ± 3.35, 5.44 ± 2.02
118
Nm으로 나타났으며, 중간 경직 그룹과 높은 경직 그룹 간의 손목 굴곡
경직 차이를 제외하고는 모든 그룹간 경직 저항 토크가 유의한 차이를
보였다. 건강한 자원자에서 우세손의 작업 공간은 액션 리서치 암 테스트
의 경우 0.53 m (좌우축) × 0.92 m (앞뒤축) × 0.89 m (상하축) 이었으
며, 일상생활동작 수행 시에는 0.71 m × 0.70 m × 0.86 m 이었다. 일상
생활동작 시에는 우세손의 작업 범위가 비우세손에 비해 유의하게 크게
나타났다. 액션 리서치 암 테스트 시 우세팔의 관절 가동 범위는 주관절
굴곡-신전 109.15 ± 18.82°, 전완 회내-회외 105.23 ± 15.38°, 견
관절 내회전-외회전 91.99 ± 20.98°, 손목관절 굴곡-신전 82.90 ±
22.52° 였으며, 일상생활동작 시에는 각각 순서대로 120.61 ± 23.64°,
128.09 ± 22.04°, 111.56 ± 31.88°, 그리고 113.70 ± 18.26°로 나타났다.
건강인과 뇌졸중 환자 간에 유의한 차이를 보이면서 동시에 뇌졸중 회복
정도에 따른 유의한 상관 관계를 보인 동작 분석 관련 변수 중, 액션 리
서치 암 테스트 4단계 검사 동작 시 전완 회내-회외의 평균 동작 크기
가 중증 기능 장애 시 건강한 사람에 비해 가장 많이 저하되었으며
(29.83%), 경증 기능 장애와의 차이도 가장 크게 나타났다 (48.46%).
영상 처리를 통한 사용자 의도 파악 손재활 로봇 시제품 사용성 평가 결
과 흥미성 (5.7 ± 1.2), 동기 유발 가능성 (5.8 ± 0.9) 및 안전성 (6.1 ±
1.1)에서는 좋은 평가를 받았으나 난이도 (4.8 ± 1.9), 편안함 (4.9 ± 1.3)
등에서는 상대적으로 낮은 점수를 받았다.
119
결론: 본 연구 결과를 바탕으로 임상적으로 유용하면서 동시에 간단하고
병상 또는 집에 휴대할 수 있는 외골격 뇌신경재활 로봇을 제작하는
데에 도움이 될 것으로 기대한다.
___________________________________________________________________
주요어: 뇌신경재활 로봇; 뇌졸중; 상지; 의공학적 인자; 외골격 로봇
학번: 2012-21741
Recommended